Even though it makes perfect sense that IQ and money would be positively correlated, the idea has encountered enormous resistance, even among some people who believe intelligence tests are excellent measures of innate intelligence. There are several reasons for this backlash:
1) Intelligence and income are two of the most highly valued traits, so suggesting a correlation between the two can be taken quite personally given the enormous income inequality in America. It’s bad enough that income helps determine our quality of life and social status, but to have it reflect our intellect too seems unfair and makes people very jealous.
2) As income inequality continues to grow, so too does resentment for the super-rich, especially in a bad economy. People would prefer to demonize rich people for being greedy and unethical than credit them for being smart.
3) Even people with limited education can get rich. This is very upsetting to people who are not rich despite attending the best schools or obtaining advanced degrees. They were told that the were superior to less educated people and are deeply invested in the Ivy League caste system, so to see some high school dropout who mispronounces words make a hundred times more money than they do, generates enormous rage. The rage can only be placated by telling themselves that the gazillionaire, despite his financial superiority, is intellectually inferior; genetically inferior. Thus the correlation between IQ and income must be denied, or at least dismissed as a statistical artifact with no direct causal implications.
4) Many of the most valuable people in history were not especially rich. Suggesting money reflects income is upsetting to the fans of such people. Although many of us understand that correlations are general trends that often don’t say much about specific individuals, when the subject is as sensitive as IQ and money, emotion trumps rationality, making it easier to just dismiss the correlation outright.
It’s interesting to observe the cognitive dissonance that occurs when you show people evidence that IQ in fact enjoys a robust positive correlation with income (+0.4). Their first instinct is to say “well maybe low IQ predicts poverty, but there’s no evidence that high IQ causes above average income.” This rationalization is win-win for the typical liberal, because it allows them to feel smarter than all the minimal wage Republicans working at Walmart without feeling dumber than the rich who they despise.
What happens when you cite evidence showing the IQ-income correlation is more or less linear through virtually the full range of IQs and virtually the full range of incomes? Then the excuse becomes, correlation does not equal direct causation. In other words, IQ causes education, and education causes income, but IQ does not cause income directly. This excuse is a win-win for the Marxist Ivy League types because it allows them to feel smarter than less educated folks no matter how rich the latter are. For they didn’t make the money the “correct” way; by going through the Ivy League gate keepers.
Such thinking helps to preserve the Ivy League caste system that elitists love because it stigmatizes the high incomes of non-Ivy League graduates as being a result of only luck, hard work, and sociopathic opportunism and denies the “free market” credit for rewarding talent. Rather it is the Ivy League that rewards talent, and the “free market” only rewards talent to the extent it rewards Ivy League grads. Thus anyone who gets rich or powerful without jumping through Ivy League hoops gets stigmatized as an enemy of meritocracy who must be destroyed.
Claiming that the IQ-income correlation is entirely mediated by education is also used by anti-HBD folks to deny IQ tests measure real world intelligence. Instead, IQ tests are dismissed as only measuring narrow test taking skills that are useful for getting a degree, but it is the degree or the test scores that are rewarded by the “free market”, not the real world intelligent behavior of high IQ people. But what does the research show? Scholar Ruth Berkowitz looked into the correlation between LSAT scores (a proxy for IQ) and income among lawyers. She writes:
The regression results show that, across the top 50 schools, LSAT scores are significantly related to starting salary, even when controlling for the cost of living in the school’s location. One point on the LSAT is worth over $2,600 on the new scale…Of course, each point on the LSAT is not equal in terms of its effects on starting salary. At the high end of the scale, one point is worth much more than it is worth on the lower end of the scale. Without controlling for any other variables, a one point increase on the new scale on the LSAT (or a 1% increase at the mean) leads to a salary increase of $3,080 (8.5%) for the top 50 schools whereas a one point increase leads to only a $1,812 (6.2%) increase for all 177 schools combined… Similarly, moving up along percentiles on the LSAT distribution brings higher returns at the high end.
So among people with law degrees, LSAT scores and future incomes are substantially correlated. But of course, not all law degrees are equal. Do LSAT scores predict income among lawyers who attend the same law school? The answer is yes:
The regression results for the individual data show that there is a significant (at the 5% level), albeit a smaller relationship between LSAT scores and starting salaries than there is for the cross-school model. Among the students in one school, one point on the LSAT is worth only about one-seventh of what it is worth in the cross-school model. These results indicate that six-sevenths of the variance is being used up in the screening effects of the school. Law schools have the ability to put more energy into screening students than do law firms. Law firms assume that in general, students attend the highest quality school into which they were admitted. Therefore, the true effect of one point on the LSAT is greater than can be measured within one school. However, in terms of lifetime income, the spread is still a significant difference even within one school. A student with a higher LSAT score, should, on average, make more money than a student who scored lower and attended the same law school. Between schools, the spread is larger. If a student scored in the top 5% on her LSAT and went to a top 5% school, she would be earning a higher salary, on average, than if she attended a lower ranking school.
So before controlling for what law school one attended, each point on the LSAT is worth $2,600, however among lawyers who attended the same law school, each LSAT point is worth only one seventh of that, so I assume $371 per year. So, let’s say we have two lawyers who attended the exact same law school, John and Ted. Let’s assume John got an LSAT score of 180 (equivalent to 152 on the IQ scale) and Ted got a score of 150 (equivalent to 111 on the IQ scale). This would predict an income difference of $11,130 dollars a year, or nearly half a million dollars over a 40 year career. So in the typical case, there starting salaries would probably look something like this:
Same law school:
Ted LSAT IQ equivalent 111: $70,000 starting salary, life time earnings over $2.8 million
John LSAT equivalent IQ 152: $81,130 staring salary, life time earnings over $3.2 million
Now a nearly half million difference is huge, but it might seem kind of small considering the two men differ by 41 IQ points. However they only differ by 41 IQ points on the LSAT. Assuming the law school both men attended was average, if they were retested on the WAIS-IV, their IQs would regress to the mean IQ of lawyers nationally (IQ 125). Because of a statistical phenomenon known as range restriction, among lawyers, especially lawyers at the same law school, the LSAT probably only correlates about 0.3 with scores on the WAIS-IV. For example I found only a correlation of about 0.3 between self-reported LSAT scores and self-reported SAT scores, and others have a found a similar correlation between LSAT scores and Bar exam scores. A 0.3 correlation means that even though John was 41 IQ points smarter than Ted on the LSAT, he would likely be only 41(0.3) = 12 IQ points smarter than Ted on the WAIS-IV.
So the bottom line is that if you took the highest LSAT student and the lowest LSAT student at every law school, their WAIS-IV IQs would probably only differ by about a dozen points, and yet their life time earnings would differ by nearly half a million in today’s dollars. So even among people with identical schooling, even small differences in IQ are associated with huge differences in money:
Same law school:
Ted WAIS-IV IQ 121: $70,000 starting salary, life time earnings over $2.8 million
John WAIS-IV IQ 133: $81,130 staring salary, life time earnings over $3.2 million
Of course it should be noted that while a 12 point IQ gap is associated with a half million difference among people who attended the same school, the same 12 point IQ gap would be associated with a $3.1 million life time earning difference when schooling is not controlled:
Different law schools:
Ted WAIS-IV IQ 121: $70,000 starting salary, life time earnings over $2.8 million
John WAIS-IV IQ 133: $147,910 staring salary, life time earnings over $5.9 million
So while people like the Lion of the Blogosphere correctly assert that where you attend college is vastly more important to your income than how smart you are, being smart is still quite important in its own right.
Now in full disclosure, I should point out that another study found virtually no correlation between standardized test scores and income among graduates of the same business school. Ronald Yeaple of authoritative Forbes magazine reports:
Our survey of hundreds of MBA alumni found no correlation between an individual’s GMAT score and that person’s post-MBA success as measured by starting salary and pay over the first five years after graduation. In this study of individual graduates, the R-square = 0.0007, which means that there was no correlation between GMAT scores and post-MBA earnings. Similar studies in the past gave the same result. Based on this research, the GMAT appears to have no power for predicting an individual prospective MBA student’s future career success. Some of the most successful managers in our study had below-average GMAT scores, and vice versa.
The explanation for this is probably that a lot of rich people use their influence to get their unqualified children into business schools they don’t qualify for, and then those same dumb trust fund babies are given a high paying job in the family business they’re also not qualified for, and delegate their responsibilities to higher IQ employees. But in a field like law, where you actually have to interpret contracts and argue cases, it’s much harder for dumb rich kids to earn high incomes.
and the relationship is NOT causal. higher IQ makes it easier to make a lot. it doesn’t cause making a lot.
This text is completely ridiculous. And i’m not a typical liberal, nor liberal or conservative cattle i’m.
you are doing a enormous confusion here.
My case, i no have envy against rich people. If i want, i already would be rich. Super rich and rich people, GENERALLY, are anti-ethic, manipulator, lack of wisdom, justice and compassion senses. I’m not rich specially because i have conscience. SOME rich people have little conscience. It is a problem of individualism, when you believe that the state will be help all of citizens.
They are smarter than MANY normal people of middle class and specially, the lower classes, but many middle class people are more REAL smart with conscience.
Many rich people suffering of Marie Antoinette Syndrome, stupid and status seeker. This people dedicate all their life to be rich or/ famous, like some people are more naturally inclined to be intelectual. It is natural, but naturality is not a excuse. Psychopathy is natural, nature tend to be very sociopathic.
Modern society is unfair because is based in a kind of “crazy race” where the agressive, selfish, manipulator will be the “winner”.
I don’t have angry or envy against rich people because i do like to be like them BUT because they are in power, then, crazy and wisdom-stupidy winner guys have total control of our lifes, it is TERRIBLE. Even if “our” government was made by ‘good’ people, inequality, when one person have all and most people have nothing is insanely amoral.
And most rich people use their power to do bad things because they no have wisdom to do best choices and/ or because they are sociopaths. If i was rich, i help other people and anima-ls, no doubt about it. But your liberal celebritches are parasites extremely hypocrites. At least conservative rich people no hide that they are primates.
Empathy is when you to put yourself in the skin of other people. If you was very poor you like to be helped. If you was very rich and real empathetic, you help other people. Most of Real smart people have aversion to money, because money is not solution, is the problem. Money buy soul of human populations, like white people today.
The high-low coalition is the problem (the coalition of the top 0.01% and the bottom 50%.) Its important that the top 50% have all the power, for various reasons to long to describe in detail here. But basically, its more stable if the top 50% have the power, and it distributes power more meritocraticly (labor providers), rather than capital providers. So just change the voting scheme.
The HL coalition works by the high bribing the low with the top 50% money, and in return they get monopolies for their capital to exploit (e.g., building developers in NYC build affordable housing as 20% of units, but the system is so complicated only the big developers can work the system, effectively granting them a monopoly.
The 20-rule for affordable housing in NYC is granted through a lottery system. If they picked you and you fit the income guidelines, then you are headed for a good-life, granted you’re one of those rare high IQ candidates who know how to exploit the system, where you mimic your well to do neighborhoods, who are paying 10x more than you in the rent, to be in the same building.
* neighbors
Very good post with more analytical content than speculative stuff.
The information from my accountant has confirmed your point in a very small circle who have the same degree and the same starting income. The one with high test score on USMLE has the highest income after ten years time.
IC, I was thinking same thing. HBD blogs have been arguing about IQ, income & education for years. Nice to finally find a blogger who understands how to do the math.
Thanks guys!
Your ‘Same Law School’ analysis will be wrong unless corrected for a concept known as restriction of range. Basically, variables are always less predictive if the range they operate on is smaller, and there exists a formula to correct for it, which roughly multiplies the predictive power by the ratio of the variances. So if law schools have 1/5th the variance as the population as a whole, the predictive power, the true predictive power, is roughly 5 times as strong as will show up. Thus, you can not attribute the decline in the variance explained to the law school selecting a way most of the variance. It might be true, but not from the evidence cited.
Here is a good discussion on restriction of range: http://pareonline.net/getvn.asp?v=14&n=5
In addition, since law schools are selecting on a variety of metrics, a higher LSAT implies lower other skills (which is not generally true, just intra law school), so the intra-law school data will be even further off. Basically, don’t bother with it unless you are going to account for these key problems first.
gs,
excellent points. I simply wanted to show that even among people with identical degrees from identical universities, there’s a positive correlation between IQ & income.
But you’re right that corrected for range restriction the correlation would be much higher
Meanwhile one can’t assume that the high correlation between income & school attended means that Ivy League degrees cause high income. I’m sure they do to some degree, but it could also be that Ivy League schools have higher income graduates because they select students who would have largely been successful anyway (because of ability, work ethic,family connections etc )
This is just more taking plain reality — which flatly disagrees with you, demonstrating that 6/7ths of the income differential is due to school attended — and trying to find ‘a way out.’
The correlation would be higher than .3. First of all, everyone who has taken the LSAT has already taken the SAT, which means that their scores on another rough IQ test wouldn’t regress much, if at all. For the most part, an individual’s law school is roughly within the same tier as their university.
Second of all, among all law schools, whatever smaller amount of variation there is certainly isn’t enough to bring the correlation down to .3. Var IQ + Var X = Var LSAT. (.81)^2 + ~.35 = Var LSAT. .81^2/5 + .35 = .48 Var LSAT. Multiplying everything by 2.08 to re-normalize gives us .27 for VarIQ which brings the correl to —> .52.
Not only that, but the .3 figure fails the ‘giggle test.’ 167, that’s 16 points above the LSAT mean, according to IQ societies, is equivalent to 130. You have also previously said that you do not believe LSAT takers are any smarter than college graduates. Fine, those two data points together mean that this new speculation cannot possibly be correct. 17*.3 = 5.1 + 113 college grad IQ = 118 != 130.
Either LSAT takers are much smarter than the college average, MENSA is just wrong, or this .3 figure is wrong. Take your pick.
‘John got an LSAT score of 180 (equivalent to 152 on the IQ scale) and Ted got a score of 150 (equivalent to 111 on the IQ scale).’
‘even though John was 41 IQ points smarter than Ted on the LSAT, he would likely be only 41(0.3) = 12 IQ points smarter than Ted on the WAIS-IV.’
Another one that fails the giggle test. 1 out of 2 people received Ted’s LSAT score or better. 1 out of roughly 4000 received John’s LSAT score. And you believe this equates to 12 IQ points. Nonsense.
The correlation would be higher than .3. First of all, everyone who has taken the LSAT has already taken the SAT, which means that their scores on another rough IQ test wouldn’t regress much,
You didn’t understand the blog post. I regressed their LSAT IQs to the mean of lawyers (not the U.S. mean) so I already factored in the fact that they have been pre-selected.
if at all. For the most part, an individual’s law school is roughly within the same tier as their university.
Citation?
Second of all, among all law schools, whatever smaller amount of variation there is certainly isn’t enough to bring the correlation down to .3.
It might not be. 0.3 was just a guestimate, but not an especially far-fetched one. The correlation between Raven IQ and SAT scores was only around 0.4 before correction for range restriction, and lawyers from the same law school are more range restricted than college undergrads at the same college
Not only that, but the .3 figure fails the ‘giggle test.’ 167, that’s 16 points above the LSAT mean, according to IQ societies, is equivalent to 130. You have also previously said that you do not believe LSAT takers are any smarter than college graduates. Fine, those two data points together mean that this new speculation cannot possibly be correct. 17*.3 = 5.1 + 113 college grad IQ = 118 != 130.
Either LSAT takers are much smarter than the college average, MENSA is just wrong, or this .3 figure is wrong. Take your pick.
You’re not understanding. 167 equals IQ 130 on the LSAT, but different “IQ tests” give different results. People who score high on one test regress to the mean on another.
Another one that fails the giggle test. 1 out of 2 people received Ted’s LSAT score or better. 1 out of roughly 4000 received John’s LSAT score. And you believe this equates to 12 IQ points. Nonsense.
No 1 out of 2 aspiring lawyers achieve Ted’s LSAT, only 1 out of 4 people from the general population would receive it if everyone took the LSAT. Both John and Ted ended up in the same occupation which means the LSAT probably dramatically overestimated the IQ difference between them given the high correlation between IQ and occupation. Thus Ted probably got unlucky on the test and John probably got lucky, so when both men are regressed to the mean of their occupation, they start to converge. The example is not equivalent to two random Americans with such scores.
Blah blah pumpkin apparently individuals who have taken the SAT and the LSAT still have some more regressing to do on an IQ test. So much regressing so as to make the correl between LSAT and IQ far less than what MENSA assumed it to be. An individual who receives two high scores likely has that equivalent level of IQ. They aren’t going to regress to any new mean. The average lawyer lsat is in the low 160’s, i.e. matches the IQ figure almost perfectly. Probably because no more or insignificant regression is going to take place.
And once again 1 out of 4000 LSAT takers received John’s score (FYI apparently I need to add “people who took the test” to get point the across) and to equate that to 12 IQ points is ridiculous. There is no other academic standardized test where over 3 SD’s equals 12 IQ points. You’re just making stuff up again.
To even say that 150 to 180 on the LSAT can even significantly be about luck (even in both directions) is to utter an absurdity.
The test retest improvement/slide is around 3 points. That’s the luck factor. 2% of time your score will be 6 points higher or lower.
Blah blah pumpkin apparently individuals who have taken the SAT and the LSAT still have some more regressing to do on an IQ test. So much regressing so as to make the correl between LSAT and IQ far less than what MENSA assumed it to be. An individual who receives two high scores likely has that equivalent level of IQ. They aren’t going to regress to any new mean. The average lawyer lsat is in the low 160’s, i.e. matches the IQ figure almost perfectly. Probably because no more or insignificant regression is going to take place.
Again, you’re not comprehending. The average lawyer’s LSAT don’t regress when they take an official IQ test because the average lawyer is what they are regressing to. It is lawyers who score higher or lower than the average lawyer who regress.
I’m very concerned that you don’t have the conceptual ability to understand this. In my opinion your IQ is well above average, but this is a very intellectual blog, and IQ 115 is not enough to discuss some of the more abstract issues we discuss here. .
To even say that 150 to 180 on the LSAT can even significantly be about luck (even in both directions) is to utter an absurdity.
The test retest improvement/slide is around 3 points. That’s the luck factor. 2% of time your score will be 6 points higher or lower.
Test retest improvement/slide? I beg your pardon. That’s not a scientific term. And luck is not just about blindly guessing. Luck also means that you got lucky that the test measured the part of intelligence you happen to have, and not some other talent (i.e. spatial ability) one might be lacking.
Pumpkin the concept you are discussing is easy to understand it’s just wrong. Whatever additional regression would take place would not be much if any. All lawyers have taken multiple tests. Their scores are accurate as is. You are saying that deviation from the lawyer mean higher or lower would imply some significant amount of luck involved in the score, hence regression. A 150 LSAT IQ according to you is actually something like 133, 25*.3 + 125 = 133.
Maybe if they’d taken one test sure. They haven’t. Whatever regression would take place would be much less.
All of my standardized tests have put me far higher than 115, so maybe I have some regressing to do, but then again I’m not surprised to hear such an assessment from you: you view a lot of nuance as hair-splitting.
This blog is very conceptual of by conceptual you mean a lot of hair-thin speculation.
And it’s rich you’re going to now “take me to task” over not using a scientific term. Perhaps it is you who does not understand scientific concepts when they are presented in ordinary language.
The amount of variance in an individual’s LSAT score tends to be in a 6 point range.
If all mental abilities correlate and the SAT and LSAT measure the most g-loaded abilities it’s not likely that these people are getting particularly lucky.
Maybe if they’d taken one test sure. They haven’t. Whatever regression would take place would be much less.
You’re still not grasping the concept. The more pre-screened they are, the less their LSAT IQ scores would regress to the general U.S. population mean of 100 but the more it would regress to the mean of lawyers (IQ 125). That’s because the more restricted the sample, the lower the correlation between LSAT and IQ and the more extreme the mean of the sample. You don’t seem to understand that less regression to a lower mean is the same as more regression to a higher mean..
100 people take an IQ test. The top 50 retake the test. If the top 50 then retake the test they aren’t going to regress to any new mean. Their scores are their scores. So we can assume that at this stage the “other” factors are mostly canceled out.
You seem to believe that the range restriction here (such as taking the top 30 of the top 50) will so radically alter the variance and correlation, inserting more randomness, so as to make the spread between high and average less than 1 SD.
In normal LSAT takers the correl between LSAT and IQ is very high. .81. If we could take a composite LSAT and SAT to IQ it’s probably higher. To get down to .3 or .4 requires a restriction that, like I said, that does not happen.
100 people take an IQ test. The top 50 retake the test. If the top 50 then retake the test they aren’t going to regress to any new mean.
The average IQ of the top 50 regresses to the mean of the 100 people no matter how many times they are tested,, but individuals within the top 50 regress to a new mean on the second retesting because if you score high in an unselected group, you regress to the mean of the unselected group, but if you score high once you’re already in a selected group, you regress to a higher mean because you scored high twice. First to get into the selected group, and then again on the retesting.
In normal LSAT takers the correl between LSAT and IQ is very high. .81. If we could take a composite LSAT and SAT to IQ it’s probably higher. To get down to .3 or .4 requires a restriction that, like I said, that does not happen.
I don’t know what the exact correlation between IQ and LSAT scores is among lawyers who attended the same law school; all I know is that it’s a heck of a lot lower than 0.81. 0.3 was just a guess based on what little data I could find.
pp has a bias that verbal IQ isn’t the real thing ignoring that it’s a better measure of the real thing than any battery of “puzzles”.
Canuckistan is 10% Asian and Asians however accomplished, however long they’ve been in the US or Canada, have marbles in their mouth.
is it genetic? God forbid! but it is a FACT.
My experience is that of all the standardized tests, the LSAT is far and away the best predictor of a candidates ability to reason through any type of problem.
The LSAT is very close to an outright IQ test, which is why suggesting such a small split between the highest score and the average score is absurd.
The LSAT is very close to an outright IQ test, which is why suggesting such a small split between the highest score and the average score is absurd.
Even on two official IQ tests, you will find very low correlations if the range is sufficiently restricted.
Yes but the range would have to be restricted in a way that is not likely the case here.
‘The average IQ of the top 50 regresses to the mean of the 100 people no matter how many times they are tested,, ‘
This isn’t necessarily true. There are many ways to view regression to the mean. I thought you agreed with godlayer’s version which iirc is autocorrelative. If you don’t then that opens up the discussion…
I don’t know why you think it’s so much lower. Law school LSAT SD’s are about 6-7 points in either direction, so 6.5^2 = 42. LSAT sd is 10 so the variance is a little less than half.
To further clarify: you are saying that we are redefining the group based on their scores on the second retest. We aren’t. It’s the same elite SAT group. So the LSAT-lawyers won’t regress.
Thus restricting the range to within law school is the only place where the correlation would be reduced.
To further clarify: you are saying that we are redefining the group based on their scores on the second retest. We aren’t. It’s the same elite SAT group. So the LSAT-lawyers won’t regress.
I was responding to your example of 100 people, but in the specific case of lawyers, first only the brightest high school students tend to take the SAT, then only the brightest of those tend to get into college, then only the brightest of those tend to take the LSAT, then only the brightest of those tend to get into law school, then only the brightest of those tend to pass law school and the bar exam.
Thus even controlling for LSAT scores, lawyers will be smarter than the average American, and for that matter smarter than the average LSAT taker. The more tests you’ve previously scored high on, the higher the mean you’ll be regressing to when you take yet another test. Got it?
Lol I’ve “gotten it” from the jump, do you? You’re not going to be regressing to a new mean. The reason you think so is because you think the group is being redefined each time.
If you took the top X% SAT scorers and gave them an LSAT it would produce the accurate scores of the lawyers-who-took-LSAT.
The group mean would be higher but there would be no further regression to it.
It’s only when you take the second score to redefine the group that they regress to the new mean, i.e. top 50 then top 30 of the top 50.
You don’t need to do that here because the group can be partitioned from the SAT without further redefinition.
It’s probably because you think the SAT and LSAT don’t highly correlate: they do. I’ve seen .89—nearly the same as test-retest.
So the only restriction you are introducing is the within law school restriction which simply would not reduce the correl by that much AINEC.
Lol I’ve “gotten it” from the jump, do you? You’re not going to be regressing to a new mean. The reason you think so is because you think the group is being redefined each time.
It is. Lawyers are more elite group than bar exam takers who are a more elite group than LSAT takers who are more elite group than SAT takers who are more elite group than the general population, thus the correlation between two IQ tests (LSAT and WAIS-IV) among lawyers is going to be much smaller than the correlation between two IQ tests in the general population, but the mean being regressed to will be much higher..
If you took the top X% SAT scorers and gave them an LSAT it would produce the accurate scores of the lawyers-who-took-LSAT.
Yes of course Swank. All you’ve done is defined the elite subset of SAT takers as those who did especially well on the SAT. By contrast, I defined the elite subset of SAT takers as those who went off to become professionals (i.e. lawyers). We’re using different measures of g (SAT scores vs occupation), but in both cases, we’ve defined a group of people from a restricted range of g that is much higher than that of the typical LSAT taker, let alone the typical SAT taker, let alone the typical American, and so within this select group, we’ll see a lower correlation between LSAT and WAIS-IV IQ, which means more regression but to a higher mean. Got it?
It’s probably because you think the SAT and LSAT don’t highly correlate: they do. I’ve seen .89—nearly the same as test-retest.
I doubt that, but that’s not the point. They could correlate perfectly,and it still wouldn’t alter the point I made above
Good you now agree you have selected a certain group of SAT takers (represented by “lawyer”). This group remains the same when taking the LSAT and when taking the bar.
So the regression occurs between SAT to LSAT, which means the LSAT score is accurate.
You are assuming I am taking a group that gets more elite each time. Once again there is no need, everyone who is a lawyer is part of ONE group.
At this point I’m not sure you “get it.” I’ve restated your point to you several times…it’s not hard to understand.
Good you now agree you have selected a certain group of SAT takers (represented by “lawyer”). This group remains the same when taking the LSAT and when taking the bar.
So the regression occurs between SAT to LSAT, which means the LSAT score is accurate.
That’s not how regression works, Swank. People who are explicitly selected based on high SAT scores (i.e. Harvard undergrads) are the ones who regress on the LSAT. The group you are describing (future lawyers) do not regress from the SAT to the LSAT because they are defined by their future occupation, not by their SAT scores, so they regress from occupation level (future lawyers) to both LSAT and SAT. In other words because lawyers are in say the top 0.5% of occupational status, they would regress to the top 5% on the LSAT, SAT, or the WAIS-IV given the imperfect correlation between occupation and test scores.
You are assuming I am taking a group that gets more elite each time. Once again there is no need, everyone who is a lawyer is part of ONE group.
And that one group has a mean they regress to, and that mean is higher than the mean of all LSAT takers or all SAT takers or the general population. That one group also has a range, and that range is smaller than those broader groups, and that range restriction results in a smaller correlation between LSAT and WAIS-IV IQ which means more regression to a higher mean, which was what I originally said.
You seem to think future lawyers score really high on the SAT and then regress to the mean on the LSAT and are arguing that they don’t regress again on the WAIS-IV. That’s not the argument. The argument is they score really high on occupation and then regress once from occupation to any of those three tests but they regress more but to a much higher mean than people who achieved lesser occupations or education levels.
‘That’s not how regression works, Swank. People who are explicitly selected based on high SAT scores (i.e. Harvard undergrads) are the ones who regress on the LSAT’
All future lawyers are indeed ‘selected based on high SAT scores.’ The LSAT scores of this group are all, nearly without exception, good enough to be ‘selected’ by law schools and become lawyers. i.e. ‘not really selected based on LSAT.’
Now, you want to use ‘occupation prestige,’ as your ‘selection’ metric. The reasoning goes, some part of the occupation prestige is luck and some part is VarIQ. And that might make sense….if you didn’t know anything about the occupation holder. Here, however, you do. If occupation is ‘score on IQ test,’ and that IQ test is administered twice, then actually we would say that the final ‘occupational prestige’ has zero luck and all varIQ. Or, as I have already said: doing what you have done is unnecessary here.
‘And that one group has a mean they regress to, and that mean is higher than the mean of all LSAT takers or all SAT takers or the general population. That one group also has a range, and that range is smaller than those broader groups, and that range restriction’
The range restriction, regression to the mean-because-of-selection-bias —- all leading to the same thing really, here — is captured in the initial SAT group sample, which is why the regression from SAT to LSAT ‘takes care of it.’
The reasoning goes, some part of the occupation prestige is luck and some part is VarIQ. And that might make sense….if you didn’t know anything about the occupation holder. Here, however, you do. If occupation is ‘score on IQ test,’ and that IQ test is administered twice, then actually we would say that the final ‘occupational prestige’ has zero luck and all varIQ. Or, as I have already said: doing what you have done is unnecessary here.
Swank obviously a person’s occupation is not 100% determined by their IQ, otherwise all lawyers would have the exact same IQ. Think about it.
All I did was used two variables to predict a person’s WAIS-IV IQ. His LSAT score and his occupation. Two variables are better than one. You seem to think that because lawyers have taken multiple tests to get where they are, their LSAT scores must correlate perfectly with their WAIS-IV score which is very strange logic
The range restriction, regression to the mean-because-of-selection-bias —- all leading to the same thing really, here — is captured in the initial SAT group sample, which is why the regression from SAT to LSAT ‘takes care of it.’
As I already explained, there is no regression from SAT to LSAT. You’re confusing future lawyers with people who scored high on the SAT. Obviously the two groups overlap, but there’s no reason to think the SAT overestimated the IQ of future lawyers anymore than the LSAT did so why the hell would they regress?. If anything, future lawyers are people who got especially lucky on the LSAT (hence the decision to study law) so any regression between tests would be in the opposite direction.
‘Swank obviously a person’s occupation is not 100% determined by their IQ, otherwise all lawyers would have the exact same IQ.’
I never said it was. I said the variance ‘occupational prestige’ is determined by the variance in IQ and ‘X.’ You saying that when we select high occupational prestige individuals, they are likely to regress on both var IQ and ‘x.’
So if you gave individuals at a certain level of ‘OP,’ an IQ test, they would regress, because essentially their ‘Var IQ’ is now broken into two parts (‘Var IQ + luck) + X’….unless….they had already taken IQ tests.
If you know that an individual with this occupational prestige level has already taken TWO tests…and you can identify those tests….you wouldn’t expect additional regression.
It’s like the point regarding the average lawyer LSAT matching the profession’s average IQ, which you took to mean something other than what it meant. The point is that, if lawyers represent a sample with ‘more than average luck’ on the LSAT, when you administered them an IQ test (which is where that data point comes from), then you’d expect their IQs to be lower on average than their LSAT scores would predict. That is not the case.
For them to regress to a new mean of 125, we would need to pull a sample from the population with a mean of 125 (lawyers), which isn’t what we’re doing —> we’re testing all the lawyers.
‘ but there’s no reason to think the SAT overestimated the IQ of future lawyers anymore than the LSAT did so why the hell would they regress?.’
The real barrier to entry is the SAT, not the LSAT. I already said that if you took a certain percentage of SAT takers, their LSAT scores — virtually all of them —- would still qualify them for ‘lawyer.’ So, the only test that would overestimate their IQs would be the SAT.
*(retested lawyers) —> point being we’d have to redefine the group afterwards, again.
Swank, you’re making it more complicated then it is. Let’s simplify.
Let’s say the correlation between IQ and occupational status is 0.7, and lawyers are +2.33 SD in occupational status.
That means, the expected IQ of the average lawyer would be 0.7 (+2.33 SD) = +1.63 SD or about IQ 125. So the regression for the entire group occurs from the occupation level to the IQ level, not from one IQ test to another.
Now is we have a specific lawyer with an IQ of 152 on the LSAT and we want to know how he’ll score on the WAIS-IV, then we should expect regression from one test to another because regression happens anytime two variables are not perfectly correlated. That’s all regression is.
So if the correlation between the LSAT and the WAIS-IV is 0.9 among lawyers, then the lawyers expected WAIS-IV IQ would be:
0.9(152 – 125) + 125 = 149
But if the correlation between LSAT and WAIS-IV among lawyers is 0.3, the correlation would be:
0.3(152 – 125) + 125 = 133
Yes, and you are inferring that the correlation between LSAT and WAIS will be less because X future lawyer’s LSAT correls less with WAIS because of the ‘future lawyer,’ (and then same law school) range restriction. You accomplish this by assuming the LSAT is the first ‘test.’
The point is that, in reality, this is untrue, because these individuals have already been tested and retested. Or in other words, you have better data points to use. Whatever the occupational prestige rank inhabited by lawyers is, it represents a group of SAT takers. This group takes another test. We have indirect evidence that this next score is stable, which means the group mean is higher but there’s no further regression.
You are expecting the Johns of the world, who received a 160 IQ on one test, then 150 on another test, to continue regressing indefinitely.
what does “test X overestimates IQ” even mean?
given a correlation <1, whatever the test, those who score high on it will, as a group, score lower on all other tests.
theoretically the one IQ, the single figure, would be the limit of the percentile score for an infinite series of tests. presumably it would converge quickly and it wouldn’t be necessary to take 1000 tests.
that is,
IQ = lim (N -> inf) of percentile(T1+T2+…+TN).
so, in my case, how are the tests i’ve scored so high on a homogeneous group rather than the beginning of such an infinite series. i keep scoring at the 99+ percentile. that seems to be my “limit”.
larding the battery with more “g-loaded” tests or measuring the first principal component rather than the sum or the mean of the tests makes no difference to the limit as long as the tests are chosen randomly. but such a series may take longer to converge.
Like anything else, regression to the mean — in statistics — is a model. You can assume infinite regression to the mean, you can assume a process where all scores after the second will reflect the second score, you can assume one where it will continue decreasing toward some point above the mean with decreasing velocity…and a lot of permutations, all of which just reflect the speed at which convergence occurs.
As I have said many times….I try to discuss with pumpkin using assumptions I have seen pumpkin make. Gs interjected once before discussing regression to the mean from one of the above-mentioned standpoints, and pumpkin seemed to agree with that standpoint. So that’s the assumption I’m working with here.
Yes, and you are inferring that the correlation between LSAT and WAIS will be less because X future lawyer’s LSAT correls less with WAIS because of the ‘future lawyer,’ (and then same law school) range restriction. You accomplish this by assuming the LSAT is the first ‘test.’
No I’m assuming lawyers who attended the same law school are a range restricted population because they’re a very specific group. In the same way I would assume Italian taxi drivers who read about philosophy and enjoy playing chess are a range restricted group.
The point is that, in reality, this is untrue, because these individuals have already been tested and retested. Or in other words, you have better data points to use.
What better data points? If all I know is an individual is a lawyer from an average law school who obtained an LSAT IQ of 152, then that’s all the data I have.
Whatever the occupational prestige rank inhabited by lawyers is, it represents a group of SAT takers. This group takes another test. We have indirect evidence that this next score is stable, which means the group mean is higher but there’s no further regression.
You have this strange idea that lawyers scored high on their SATs and then regressed to the mean on the LSAT. The reason you think this is because a random group of high SAT people would regress if given the LSAT (or any other test) but that’s because random high SAT people scored high, partly for random reasons. Lawyers are not a random group of high SAT people anymore than they’re a random group of high LSAT people or a random group of high WAIS-IV people, thus their scores on all three tests should average about 125.
You keep talking about the entire occupation regressing but I’m talking about individual outliers within the occupation regressing.
You are expecting the Johns of the world, who received a 160 IQ on one test, then 150 on another test, to continue regressing indefinitely.
If all we know about John is that he’s a lawyer from an average law school who scored IQ 152 on the LSAT, then the best estimate of his WAIS_IV IQ is that it will be in between his his LSAT IQ and the average WAIS-IV IQ of lawyers from his law school.
And incidentally, the same logic would apply to estimating what his SAT IQ was back when he was in high school. It too should be in between his LSAT IQ and the average SAT score of lawyers who attended his law school. How far in-between depends on the correlation between the LSAT and the SAT among lawyers from his law school.
You seem to think the order that people take tests has anything to do with it. It doesn’t.
what does “test X overestimates IQ” even mean?
It means the score was higher than one usually scores
given a correlation <1, whatever the test, those who score high on it will, as a group, score lower on all other tests.
Correct
‘No I’m assuming lawyers who attended the same law school are a range restricted population because they’re a very specific group’
I know they are, and what I have said this entire time, is that that is the only restriction of range that would occur in this group. The restriction of range from lawyers generally to lawyers in the same law school is not as great as you imagine it to be.
‘You keep talking about the entire occupation regressing but I’m talking about individual outliers within the occupation regressing.’
I know, and the point about the IQ score matching the LSAT score of lawyers is that there doesn’t seem to be a lot of ‘luck’ involved in any particular score. The group of LSAT takers that become lawyers should regress further, collectively, if these scores do not represent the ‘true score,’ when you give them an IQ test. They don’t.
‘You seem to think the order that people take tests has anything to do with it. It doesn’t.’
Nope. I think the number of measurements on the trait we have has ‘anything’ to do with it. It does.
‘If all we know about John is that he’s a lawyer from an average law school who scored IQ 152 on the LSAT, then the best estimate of his WAIS_IV IQ is that it will be in between his his LSAT IQ and the average WAIS-IV IQ of lawyers from his law school.’
Yes, but we know more than that. John is a lawyer with 1100 on his SAT then 152 on his LSAT. His next score will probably be the equivalent of 152 LSAT IQ on whatever other test. The only restriction that occurs is when you limit the sample to ‘same law school.’ The ‘same law school’ restriction is not nearly enough to lower the correlation to .3.
I know they are, and what I have said this entire time, is that that is the only restriction of range that would occur in this group. The restriction of range from lawyers generally to lawyers in the same law school is not as great as you imagine it to be.
The range restriction of lawyers generally is itself a lot. These are people who not only graduated from college, but got accepted into law school, passed law school, and then passed the bar exam.
‘
I know, and the point about the IQ score matching the LSAT score of lawyers is that there doesn’t seem to be a lot of ‘luck’ involved in any particular score. The group of LSAT takers that become lawyers should regress further, collectively, if these scores do not represent the ‘true score,’ when you give them an IQ test. They don’t.
That’s because collectively, lawyers were not especially lucky on the LSAT. The lucky ones and the unlucky ones cancel each other out so their average LSAT IQ accurately reflects their true intelligence. But individuals lawyers can be extremely lucky or unlucky on the LSAT or any other one test.
Now if you were talking about a really selective law school like Harvard Law or Yale Law, then you should expect to also see collective regression though regression to an even higher mean than that of lawyers generally..
Yes, but we know more than that. John is a lawyer with 1100 on his SAT then 152 on his LSAT. His next score will probably be the equivalent of 152 LSAT IQ on whatever other test.
John was just the name I gave to the highest LSAT scoring lawyer in an average law school. We don’t know that person’s SAT score, but if it were 1100 (IQ 114), then we would not expect his IQ to be 152 on the next test he took, just because his LSAT IQ was 152. This would be an example of multiple regression where two variables (LSAT IQ) and (SAT IQ) are regressed to the mean of lawyers at John’s law school to predict a third variable (WAIS-IV IQ).
The only restriction that occurs is when you limit the sample to ‘same law school.’ The ‘same law school’ restriction is not nearly enough to lower the correlation to .3.
As I said from the beginning, 0.3 was just an educated guess on my part but I haven’t thoroughly investigated it; the true figure could be more than double that. But the point is, the individual lawyers at a specific school would regress to the mean of their school on a second testing…by how much depends on the correlation between the two tests in that restricted population
‘That’s because collectively, lawyers were not especially lucky on the LSAT’
Either that or the LSAT scores are more or less the true scores. And that’s strange to consider the top part of a test’s distribution as collectively not having been ‘lucky.’ But, if you think an individual who takes an IQ test twice won’t have his true score, then ok.
‘We don’t know that person’s SAT score, but if it were 1100 (IQ 114), then we would not expect his IQ to be 152 on the next test he took’
I said 152 LSAT….
‘But the point is, the individual lawyers at a specific school would regress to the mean of their school on a second testing…by how much depends on the correlation between the two tests in that restricted population’
I didn’t argue against this on its own. The same principle would be to test this restricted group two more times. Do you believe the second retest scores would also regress? It seems as though you do.
And that’s strange to consider the top part of a test’s distribution as collectively not having been ‘lucky.’
No it’s not strange. The fact that they went on to pass law school and the bar exam means their high collective scores were not luck. They’re not the equivalent of random high LSAT people.
By contrast the individual lawyer who scores way higher than other lawyers has not necessarily done anything to distinguish himself from other lawyers, other than scoring higher on one test, so his score likely does include substantial luck
But, if you think an individual who takes an IQ test twice won’t have his true score, then ok.
The fact that he has done well on an IQ test in the past gives more information to predict the true score. That’s the point of regressing to the mean of lawyers and not the general population. Lawyers are a group of people who have done well on tests in the past so regressing to their mean adds accuracy.
I didn’t argue against this on its own. The same principle would be to test this restricted group two more times. Do you believe the second retest scores would also regress? It seems as though you do.
I don’t care if they’re tested a trillion more times. An individual lawyer’s IQ on any one test he’s taken will (on average) regress to the second test mean of lawyers from his school for any other IQ test he’s taken..
I’m not saying someone will score 160 on the first test, and then 150 on the second, and then 140 on the third, but what I am saying is that any one of the tests one has taken can be regressed to the group mean to predict scores on any other test taken.
‘No it’s not strange.’
In this group, those other achievements have almost no cognitive component. The component that matters ‘work ethic,’ won’t really make a difference on a new IQ test.
‘An individual lawyer’s IQ on any one test he’s taken will (on average) regress to the second test mean of lawyers from his school for any other IQ test he’s taken.. ‘
So then your answer is yes? Lawyer from same law school with a pre-178 LSAT takes another LSAT, which regresses toward the mean of the law school. Then, he takes another LSAT…his score will keep regressing ‘a trillion times,’ according to you?
In this group, those other achievements have almost no cognitive component. The component that matters ‘work ethic,’ won’t really make a difference on a new IQ test.
Performance in law school has a substantial cognitive component; that’s why such students are screened with the LSAT in the first place.
‘So then your answer is yes? Lawyer from same law school with a pre-178 LSAT takes another LSAT, which regresses toward the mean of the law school. Then, he takes another LSAT…his score will keep regressing ‘a trillion times,’ according to you?
A student’s likely score on test 2 regresses from his actual score on test 1 (or any other test) and a student’s likely score on test 3 regresses from his actual score on test 2 and a student’s likely score on test 4 regresses from his actual score on test 3 but we don’t get infinite regression because likely scores only regress from actual scores, not other likely scores.
You can’t predict from a prediction.
Got it?
In the group who become lawyers, the difference is mostly if not all, down to work ethic.
So your answer is yes, you believe the “likely” score would hypothetically regress forever.
So if there were two actual high scores on an IQ test, you would say the third would likely regress.
Contradicts your earlier belief or agreement with said belief, got it?
the regression of scores on tests 1 and 2 gives you a conditional distribution for the score on test 2 given the test 1 score. from this distribution can be deduced the score on test 1 and the test 1 test 2 correlation, so when one uses the test 2 test 3 correlation to find the conditional distribution for test 3, it’s just the same as that derived from the test 1 test 3 correlation and the test 1 score.
that is, you can regress on likely scores as long as the likelihood distribution isn’t just derived from the correlation.
if i said the score on test 1 was x and the score on test 2 was y1 with probability p1 or y2 with probability y2, this could be used to derive a predictive distribution for the score on test 3.
In the group who become lawyers, the difference is mostly if not all, down to work ethic.
If everyone who even takes the LSAT were accepted into law school regardless of their LSAT scores or undergrad grades, those who actually graduated and passed the bar exam would be expected to have substantially higher LSAT scores than the general LSAT population, so lawyers are not equivalent to a random group of high LSAT people. The latter enjoyed substantial luck; by contrast passing law school and the bar is independent evidence of high IQ meaning their high LSAT scores were not just a fluke.
So your answer is yes, you believe the “likely” score would hypothetically regress forever.
It depends what you mean by “regress forever”. The term implies each successive test one takes will get closer and closer to the same mean. No I don’t believe that at all because the probability of a predicted score occurring depends on the certainty of the score it’s being predicted from. A prediction based on an actual score is likely because it’s based on certain data, but a prediction based on a prediction based on a prediction becomes less and less likely with each successive prediction because the uncertainties add up, But if all we know about an individual is his score on any one test he took, then his score on any other test he took or will take will regress to the mean of the group he belongs to.
So if there were two actual high scores on an IQ test, you would say the third would likely regress.
If all we know about an individual is his score on test 2, then his score on test 2 will regress to the test 3 mean of the broader group he belongs to when he’s tested on test 3, but if we also know he scored high on test 1, then when tested on test 3, his score on test 2 will likely regress to the test 3 mean of the subgroup who scored high on test 1. Capiche?
Contradicts your earlier belief or agreement with said belief, got it
I assume you’re referring to my belief that second generation immigrants regress to the mean but the third generation doesn’t. In that case first generation immigrants are an especially intelligent subgroup of Third Worlders so the second generation regresses to the genetic mean of the broader group, but second generation immigrants can not be an especially intelligent subgroup of themselves so no further collective regression occurs in the third generation of immigrants.
But an individual third generation immigrant who takes an IQ test will likely regress to the mean of third generation immigrants on another IQ test he takes or has taken.
‘ by contrast passing law school and the bar is independent evidence of high IQ meaning their high LSAT scores were not just a fluke.’
The selection for ‘passing law school’ and ‘passing the bar’ on any individual with an LSAT score above maybe 140-145 is not a ‘cognitive sort.’ It is hard to distinguish yourself in law school — in both directions.
‘If all we know about an individual is his score on test 2’
‘a prediction based on a prediction based on a prediction becomes less and less likely with each successive prediction because the uncertainties add up’
This is if you take it out to be .correl between test-test^x(score on test 1).
More importantly, you admit that if we define our ‘broader group’ on test 1, that by test 3, the individual’s likely score would NOT regress to that group’s mean. Instead, it would regress to another, redefined group within the broader group. So, if we kept the group parameters the same, by test 3, no or little regression would occur to the original group’s mean? Do you agree?
‘second generation immigrants can not be an especially intelligent subgroup of themselves so no further collective regression occurs in the third generation of immigrants. ‘
So….if we have three tests given to a group called ‘future lawyers,’ do you agree that on test 3, an individual’s score would not regress to the mean of this ‘broad’ group?
The selection for ‘passing law school’ and ‘passing the bar’ on any individual with an LSAT score above maybe 140-145 is not a ‘cognitive sort.’ It is hard to distinguish yourself in law school — in both directions.
There’s a moderate correlation (0.3-0.4) between LSAT scores and law school grades and passing the bar exam and if everyone who took the LSAT all got into law school and all attended the same law school, those correlations would be higher still.
So, if we kept the group parameters the same, by test 3, no or little regression would occur to the original group’s mean? Do you agree?
It depends what the group parameter is and it depends what score you’re using to predict their score on what test. If the group is defined as people who scored high on the first test, and if you know only a group member’s score on the second test, then on the third test, he would likely regress to whatever is the mean on the third of those who scored high on the first test. Got it?
So….if we have three tests given to a group called ‘future lawyers,’ do you agree that on test 3, an individual’s score would not regress to the mean of this ‘broad’ group?
They wouldn’t regress to the broader mean on any test, not third test only.
You’ve already seen the comment.
I mean the sample of students in danger of failing may be a little higher or lower. The curve requires 5% C’s. So an individual need only perform better than 3 others in a class of 60 to pass. The LSAT simply does not meaningfully explain this low hurdle.
I challenge the anti-HBD crowd to demonstrate why they think that income (that is, SES) and IQ are not causally linked in a genetic sense? SES definitely doesn’t have any environmental effects on IQ since in twin studies and GCTA we find no shared-environment effects.
This is the definition of shared environment: “aspects of the environment that people living together will share.” This is calculated by first squaring the correlation between twins, adding the remainder correlation squared to that result, then subtracting that from 1, to get the ‘shared environment.’ Of course, assuming that the full correlation between twins represents the whole of a genetic contribution is faulty. To respond, HBDers say ‘look at the similar numbers with twins raised apart,’ which in reality have twins sharing many aspects of their environments —- same town, similar school, etc. etc. For example, one twin going to Middle Class School and another twin going to Middle Class School 2 would be ‘non-shared environment,’ even though in REALITY, it pretty much is shared environment.
and in fact a real scientist, Kempthorne, commented on this saying:
What, indeed, is the “grip” on environment in the human IQ area? It is no more than “reared together” versus “reared apart”, and what does “reared apart” mean? Nothing more than at some age two related individuals, e.g., identical twins or full-sibs were separated by adoption, and then placed in homes that could be related familially and/or, of similar economic and social nature. I can only comment: Really, how naive can one be? The Burt study was characterized in the literature as the “only experiment”. Some experiment!
As soon as one turns to any behavioral measurements, the need to incorporate intra-uterine, family and community environment is obvious. I have the view that the “hereditarians” are utterly naive. It is obvious that parental IQ influences offspring environment. It is obvious that there is cultural transmission. To ignore the existence of this is merely stupid. I see no point in mincing words. If non-scientists sometimes have scorn for some supposed “scientific work”, they should not be faulted.
To this naivete in model formulation, must be added a statistical naivete. Any statistical test has, within its conceptual underpinning, a sensitivity or power function. It needs essentially no deep thought to realize that the sensitivity of statistical tests for maternal effects, for genotype-environment interaction, etc., etc., is so low that to say, as has been said often, that such and such a model modification has been examined and found unnecessary is utter naivete.
the truth hurts pp.
and now pp is spouting this nonsense she undoubtedly imbibed from one of her high IQ society moron friends that only tests like Hoeflin’s mega test can measure IQs above 135.
Hoeflin is a moron. and his tests are retarded.
remember pp in the Esquire which profiled Hoeflin and his merry band of morons, Alan Guth was also profiled. the point, pp, was that Guth was actually smart and Hoeflin and co were stupid.
destructure is like you, he actually thinks high IQ society members are smart. the aren’t!
spitzer, stiglitz, bernanke, krugman, all jews, is what a 160 IQ looks like. not that impressive really.
there’s no such thing as “shared environment” Kev. it’s a fiction of the additive model. model != reality.
Hah! You make me laugh. You and swank to this day have no answer to the decades of research (http://www.udel.edu/educ/gottfredson/30years/Rushton-Jensen30years.pdf) put out by Rushton and Jensen.
Here’s yet more that anti-HBDers can’t answer to: http://unqualified-reservations.blogspot.ca/2009/01/gentle-introduction-to-unqualified_22.html
Oh, and let’s not forget that you fools have no answer yet to Cochran’s / Harpending’s NHAI yet.
you’re too dumb to talk to Kev.
the genuine producers rarely make more than 120k per year. that’s engineers, scientists, and tradesmen…the people who make the modern world.
how can that be? isn’t intelligence the ability to get things done? to produce? if it isn’t why should it be valued?
it can be because high pay in the modern world comes from stealing and exploiting not producing.
how can that be? isn’t intelligence the ability to get things done? to produce? if it isn’t why should it be valued?
it can be because high pay in the modern world comes from stealing and exploiting not producing.
Intelligence is also one of the greatest sources of evil. Ever heard the term evil genius? No one has done more to ruthlessly exploit and steal from the planet than humans and we’ve done so precisely because we’re the smartest animal. And the smartest human populations generally have killed the most people precisely because they had the technology to do so.
Intelligence is valued because it’s the mental ability to turn situations to our advantage. If we’re good people, it’s to our advantage to use our intelligence to help others because that makes us feel good. If we’re bad people it’s to our advantage to hurt others because that makes us feel good.
Intelligence is simply the mental ability to figure out how to reach our goals. They can be good goals or bad goals. For most people, the goal is money or the infinite advantages money buys, thus intelligence correlates with income
you’re too dumb to talk to Kev
And a bunch of us feel the same way about you. I estimate your IQ to be 135, which is incredibly high, but not high enough for the topics you’re most interested in.
You’re certainly not in the same league as hardcore intellectuals like professor Hsu who I estimate to be around 154.
a bunch of us?
you’re IQ is only 45.
not only do stupid people not know they’re stupid, they think that those who are smarter than they are dumber.
for example, every single member of high IQ societies have low IQs.
so if you “would estimate my IQ blah blah blah” it must be MUCH MUCH MUCH higher.
i have scored higher on IQ tests than Steve Shoe, yet hos IQ is higher than mine. pp’s estimates are better than the real thing? what a fucking joke.
you’re a moron pp.
for example, every single member of high IQ societies have low IQs.
There’s a mix of people in high IQ societies. Some would regress precipitously to the mean if given conventional tests, others are literally FIFTY POINTS above your level and it would be completely obvious to anyone in less than five minutes.
i have scored higher on IQ tests than Steve Shoe, yet hos IQ is higher than mine. pp’s estimates are better than the real thing? what a fucking joke.
I have observed you in an intellectual environment almost every day for many months and have seen you discuss a wide variety of subjects. I’ve been lucky enough to correspond with some of the greatest minds in the world and so I’m a very keen judge of the difference between a good test taker and a genuinely high level of g. You claim to have scored extremely well on college admission tests and such tests are excellent measures of g, up to about IQ 135. Beyond IQ 135 you get a ceiling bumping effect. You would do excellent on tests of general knowledge, vocabulary, and simple math, but I don’t think you could solve a truly complex and truly novel abstract problem.
Lol more of these subjective assessments….
asians are at 14% in Canuckistan. i had no idea. they’re the Mexicans of Canada.
so why is the CAD at 79 cents (.79 USD)?
subjective assessments are a huge problem. you can get away with anything, and even more, the grader can convince himself he’s objective. it’s happened to me too many times to remember.
subjective assessments favor the rich and well connected and those with “team spirit” etc. they are biased by definition.
i scored at the 99+th percentile on the GMAT verbal section which does not test vocabulary btw but has a few reading comp questions and a lot of “writing correction” questions. but i scored in the 6th percentile on its analytic writing. the 6th percentile! how can that be? at least 6% of GMAT takers have English as a second language. right? easy. stupid people think smart people are stupid.
that’s why i don’t like affirmative action and quotas…unless they’re objective. so when a rich black kid scores the same as a poor white kid admit or hire the poor white kid. but when the backgrounds are similar it would be reasonable to prefer the black.
Mercastan can’t confront its class issue. as a result it has become a third world country.
High IQ individuals are mostly moral but IQ is the root of true evil
Lol pumpkin…if high IQ moral people so vastly outnumber the high IQ immoral people why is the world as thus?
pumpkinhead–
Studies show that people do an excellent job of estimating those with lower IQ but not so well at estimating with those with higher. When Ross first showed up on my blog he claimed a certain SAT score assuming it was higher than mine. It wasn’t. To be honest, I’ve seen no evidence he scored what he claims. He fancies himself some quirky genius which is why he’s always trying to drop obscure and off-the-cuff references. Anyone who reads it can see he’s full of smoke… and mentally ill.
pp, no legit IQ test can measure above 160. i hit the ceiling on the GRE. what’s that?
so yeah there are some who claim to have IQs of 210, but they’re idiots. because there are no such IQs.
i’m NOT a genius at anything…obviously ;)…but i combine a talent for math and natsci with a high verbal IQ. people with high V and ok M or high M and ok V are a dime a dozen. those with both high are rare.
1.high score on old exam 100 of the SoA.
2. 99th percentile on all four of the American Chemical Society’s exams and on the GRE subject test in chemistry.
3. represented my uni at College Bowl.
the guy on my College Bowl team now works as a securities lawyer in Moscow for one of the five “magic circle” law firms. i answered as many as him at college bowl. but he was a russian major and i was a math major, so he likely would not have hit the ceiling of the GRE. the secret of scoring super high is being well-rounded not being gifted at any one thing which i am not.
destructure is lying as usual.
he did NOT score above 1560 on his SAT.
obscure and off-the-cuff references.
wtf? obscure? really? wasn’t my intention!
look at my f—ing posts! they’re full of typos. i type them as fast as they come. that’s it. there’s no artifice or pretense in them at all.
so you see d’s post confirms what i’ve said. and that he did not score >=1570 on the SAT. i qualified for the BGI study with my SAT scores. d hasn’t, because he can’t.
Studies show that people do an excellent job of estimating those with lower IQ but not so well at estimating with those with higher. When Ross first showed up on my blog he claimed a certain SAT score assuming it was higher than mine. It wasn’t.
He claims to have scored 1560 which on the old SAT, equates to an IQ of 160 (high enough for Prometheus). I estimate his true IQ to be 135 but I’d be curious what you think the correct number is.
To be honest, I’ve seen no evidence he scored what he claims. He fancies himself some quirky genius which is why he’s always trying to drop obscure and off-the-cuff references. Anyone who reads it can see he’s full of smoke… and mentally ill.
He knows a little bit about a lot, which takes some g, but he lacks a deeper understanding. Jensen would have described him as having high breadth, but low depth.
destructure is definitely very bright.
He’s been able to grasp certain concepts that have baffled most HBD commenters for years such as how IQ and money can be correlated above 120 even though most millionaires are not above 120 (see his comments on Lion’s blog). He also understood why there’s a correlation between how high a life form is on the evolutionary tree and how advanced it seems. Most people just deny the correlation, rather than come up with a coherent explanation.
pumpkin those aren’t concepts that have baffled anyone…they’re just groundless speculation that happen to conform to a certain stereotypical paradigm.
also the last bit of speculation is silly. ‘higher up’ on the evolutionary tree? it’s probably because most people (HBDers included) don’t understand evolution. Evolution = survival and reproduction. Insects and bacteria would be the ‘highest’ evolutionary species, on evolution’s terms.
pumpkin those aren’t concepts that have baffled anyone…they’re just groundless speculation that happen to conform to a certain stereotypical paradigm
It may very well be groundless speculation but the arguments people have made against it showed conceptual confusion. People argue that since most successful people have IQs in the 120s, therefore IQ doesn’t predict success above 120. As destructure correctly understood, people with IQs in the 120s outnumber people with IQs in the 150s so of course they’re going to be more numerous at all levels of success, including the most elite levels. But IQ 150s could still have greater individual odds of becoming super successful than 120s do.
also the last bit of speculation is silly. ‘higher up’ on the evolutionary tree? it’s probably because most people (HBDers included) don’t understand evolution. Evolution = survival and reproduction. Insects and bacteria would be the ‘highest’ evolutionary species, on evolution’s terms.
And destruction would be the first to agree that speaking of “higher up” is silly, but at the same time, evolution forms a tree, and some branches are higher than others. destructure understood why those on the “higher” branch seem more evolved to those who subscribe to a hierarchical progressive view of evolution.
If the mean between those who make 100k and 1 gazillion doesn’t significantly change, that indicates a low correlation.
In education with a correl of .65 we see the mean significantly jump and climb with each level.
In income we also see the mean jump and climb until a certain (low) level.
It can’t be explained away by “there’s just more of them.”
You could check proportions…what proportion of 150s become gazillionaires versus 120s.
People who take a progressive hierarchical view of evolution view intelligence or complexity as the trait that determines evolutionary progress.
the ultimate problem here is a cognitive bias. http://en.wikipedia.org/wiki/Cognitive_bias
specifically a version of the eco fallacy.
it’s the idea that someone’s external traits are always indicative of his internal traits. that is, a person’s situation in the world and in his particular society always follow from some essential traits, psychological traits or physical traits.
but in fact what one finds on examination is that for any given situation there are a myriad of personalities, but some types are more common than they are in the population at large.
and of course the very idea of essential characteristics of the individual is dubious. that is, who i am and who you are depends to some appreciable extent on the societies to which we belong.
the Stoics had it right. one should “take responsibility” for his representations and judgements, but he should be indifferent/apathetic to his external situation.
the equating of external situation with who one is is the height of ideology and quintessentially American, that is,, vulgar and satanic.
If the mean between those who make 100k and 1 gazillion doesn’t significantly change, that indicates a low correlation.
In education with a correl of .65 we see the mean significantly jump and climb with each level.
In income we also see the mean jump and climb until a certain (low) level.
In income, the relationship probably looks something like an 8 IQ point jump for every ten-fold increase in financial success:
Three figure income earners (the homeless): Average IQ 84
Four figure income earners (part-time minimum wage): Average IQ 92
Five figure income earners (middle class): Average IQ 100
Six figure income earners (future millionaires): Average IQ 108
Self-made deca-millionaires: Average IQ 116
Self-made centi-millionaires: Average IQ 124
Self-made billionaires: Average IQ 132
Self-made deca-billionaires: Average IQ 140
Now holding education constant, I suspect the jumps would be roughly half as large on average, but it would vary depending on the level and type of education being held constant.,
…always follows from…
Aurelius, a Roman emperor:
But a brief existence is common to all things,
and yet thou avoidest and pursuest all things as if
they would be eternal. A little time, and thou shalt
close thy eyes; and him who has attended thee to thy
grave another soon will lament.
it is still the case that the rich live longer than the poor, but not nearly to the extent they once did.
iirc a 16th c English aristocrat had a life expectancy > 70 if he lived to 20.
Forbes once reported that members of the Forbes 400 had a greater life expectancy than that of the US population as a whole. the obvious problem with this figure is that it often takes living a long time to accumulate a billion or more. that is, just getting on the list requires being pretty old.
but even then the mean age at death was reported as only three years greater then that of the average American male (it was usually a male)
Sam Walton died at 74, and one of his sons died in a plane crash in his 50s. the Waltons are worth almost 200b today.
If the correl is .4 and being dumber causes lower income, someone making three figures should be much much dumber.
Of course the average homeless person makes more than that. Panhandling for even 3 dollars a day gets you 4 figures.
Part time minimum wage also makes 5 figures. 27/40*2080*7.25= > 10k. Could be a little more or less, but the overall historical average for part-time workers is > 25 hours per week.
These jumps (ten fold increases) are much larger than the education jumps too. So the correl all the way up to the top could be even lower.
Swank, I explain where I got the 3 figure value here:
https://brainsize.wordpress.com/2014/05/18/iq-and-income/comment-page-1/
GROS PD VA
i have a big booty like alexis texas
and pp’s hero has said as much.
high IQ and genius are two different things. but some forms of genius have high IQ thresholds.
so Jensen agreed with Gladwell and disagreed with prof shoe.
in general people with high IQs are not particularly good at anything and lead fairly ordinary undistinguished lives.
no pp the correct interpretation is that:
1. law firms hire based on LSAT and law school grades and above all on the law school attended and law schools select based partially on LSAT. your data is meaningless.
2. the GMAT verbal is not very g loaded imho, but getting a perfect score on the GMAT quantitative section is actually quite rare.
and the data conforms to my, Lion’s, and Swank’s opinion and the TRUTH. that past a certain not very high threshold there is no relationship between IQ and income.
1. law firms hire based on LSAT and law school grades and above all on the law school attended and law schools select based partially on LSAT. your data is meaningless.
Law firms are not supposed to ask about LSAT scores and it’s considered extremely offensive when they do, though some break this taboo and do so anyway, much to the horror of the job applicants.
and the data conforms to my, Lion’s, and Swank’s opinion and the TRUTH. that past a certain not very high threshold there is no relationship between IQ and income.
Actually that’s not what Lion currently believes (though he may have believed it in the past). He believes that for college grads with the same degree from the same school, IQ does not determine how much money you make, unless you become a gazillionaire, in-which case he believes having a super high IQ increases the odds.
Only take Lion’s beliefs with a grain of salt. He has no life experience for more than 1/2 of his assertions.
JS, I thought it was funny when you got mad at Lion for doing yet another post trashing Staten Island. LOL!
My advice, which is better than Lion’s, when it comes to someone who has high IQ and don’t have the connections to a prestigious career. A high IQ person needs mimic someone who is a low IQ, in order to draw gov’t subsidies and take it from there. Don’t waste your time grinding at a practical, yet dead end job, and not self actualizing. It’ll make your miserable!
JS, I can’t say I’m in a self-actualizing career because my career has nothing to do with my passions for IQ and horror films, but I love my job nonetheless because I get to drive all over the city meeting fascinating and important people and I have a great deal of autonomy.
But different cultures have different values. Even high IQ Canadians (particularly in small towns) often don’t care about elite status, they just want to be part of the community and do good. The other day I met a brilliant computer programmer who makes great money, yet moonlights removing snow from everyone’s driveway because that’s the town he grew up in, and wants to give back. He and his wife were unbelievably happy people
yeah it’s pretty clear pp’s job is crap.
what do you produce pp?
‘Law firms are not supposed to ask about LSAT scores and it’s considered extremely offensive when they do, though some break this taboo and do so anyway, much to the horror of the job applicants.’
The law firms are indirectly selecting based on LSAT score by recruiting from school X or Y. The odds of getting an applicant with good law school grades who has an LSAT < school's 75th percentile are low.
Yes, we understand that Swank. That was the whole point of the post. Controlling for the law school attended.
Thus anyone who gets rich or powerful without jumping through Ivy League hoops gets stigmatized as an enemy of meritocracy who must be destroyed.
I don;t think this is true. What we find instead is that those of modest intellectual means are more likely to resent the elite, including the cognitive elite. The Ivy Leaguers are much more frustrated at the dumbing down of society by people of modest IQs and modest incomes than they are frustrated at the elite, rich people who didn’t go to an Ivy League.
What we find instead is that those of modest intellectual means are more likely to resent the elite, including the cognitive elite.
Unfortunately they don’t seem to resent the Ivy League elite, as evidenced by the fact that they keep voting Ivy League grads into the white house. But then the Ivy League media shoves them down the public’s throat
if by IQ pp means WAIS score then she might be right, but why would she mean WAIS score and not SB score or Raven’s score or SAT or LSAT or AVB or whatever?
to speak of IQ as one thing or one figure shows you have no idea what IQ means.
My decision to define IQ as WAIS IQ is arbitrary but the point is, when measuring the IQ of law students, you don’t want to use the LSAT because law students were selected based on the LSAT so statistically, their LSAT scores are going to behave differently than their scores on any other IQ test.
For example, if you looked at the LSAT IQ of Harvard Law students, you would find it to be higher on average than their IQ as measured by any other test, because their scores on every other test would regress, causing the LSAT to be an outlier (for them).
Similarly, if I was looking at the IQ of Harvard undergrads, the SAT would be inflated, but for them, the LSAT would give good results.
If I were interested in the IQ of the Mega society, I wouldn’t use the Mega test, because they did well on that test by definition, I would use the LSAT, or the SAT, or the WAIS, or whatever other test they took.
If I were studying the IQ of Jeopardy Champions, I would not use a general knowledge test to measure their IQs, even though for other people, general knowledge is a strong proxy for IQ
My personal theory is never measure the IQ of a group of people by the test used to select those people, and never correlate that test with any other variable, because the test that was used to select them, will be especially range restricted for them
If everyone in the Mega society scored in a similar range on another test, you’d say that using the other test would be a good measure of their IQ?
that should’ve been ASVAB obviously.
that’s a test given to military recruits in the US.
Pumpkin Person, please read and comment about this article by Scott Alexander: http://slatestarcodex.com/2015/01/31/the-parable-of-the-talents/
It’s about someone’s personal experiences in the light of HBD related issues, such as intelligence. Perhaps you can make a post in the same vein about your own experiences? That’d be amazing to see.
the obvious truth is that:
1. conservatives are either stupid or evil.
2. so are straw man liberals…that is, conservatives actually believe that their opponents are something they are NOT. that is, there are hereditists, but in fact there are no “environmentalists”.
3. whatever the system an appreciable fraction is going to be screwed by it. therefore however good it is there has to be a real social safety net.
4. Darwin said: If the misery of the poor be caused not by the laws of nature, but by our institutions, great is our sin. uhhh…Darwin douche…modern life isn’t Nature. it isn’t now and wasn’t in your day. the modern world is man made. even if the poor were stupid or lazy, that is NOT a good reason for anyone to be miserable…especially if their misery can be relieved by the smart and industrious.
from each according to his ability…? “how could Lenin possibly have thought people differed in ability?” asks the conservatard.
the “liberals” that conservatives talk about DO NOT EXIST!
smart people, that is all scientists (psychologists are NOT scientists), think HBDers are stupid…
because they are!
Pumpkin I like your blog but your IQ estimates are way too generous. No way could Swanknasty score 115 on a profession IQ test. That’s Ashkenazi Jewish level. 100 is a more realistic guess.
Save the fact that I already have scored much higher….
But I’ve always enjoyed subjective assessments and what people think they know.
And if it makes you feel any better, I don’t think those tests or standardized tests measure much of anything innate. So maybe I’ve just successfully bluffed my way through all of them, my entire education, etc. etc. Anything is possible 🙂
Why do you think, P.P., that early IQ testers were seen as racist and Social Darwinist?
Because they found IQ differences between races and classes and claimed those differences were genetic. People consider that racist even though it’s not; but it is social Darwinist.
The tl;dr is you think we only know one score because you posited or put forward one score. You think that we have defined the group as “high scorers” on another test, then tested that group again, which is why you believe another test will result in a regression toward this group’s mean.
But if this group were defined before that first test, they would have all been tested on test 1. The next round of scores would regress toward the mean of this group. The third would not, unless we further redefined the group (same law school).
You keep saying “if we only know one score,” in response. This isn’t true, we know two.
It doesn’t seem like you get it, which is rich.
Swank, anything you get, I got at age three. I defined the group as lawyers at an average law school from the jump. Implicit in that definition is that they’ve generally done well on multiple tests, but since you kept obsessing over previous tests, I tried to frame in terms of that, but this just confused you more and your comments are becoming increasingly strange..
I think it might be better if you just posted on the Lion’s blog for the next month or so.
And at the jump I said the only restriction here was the same law school, when it came to the LSAT being less accurate.
Implicit in the definition is doing well on tests relative to others not themselves. If we know their performances relative to one another on two tests we have no reason to think that the next score of an individual will further regress, in my words, much if at all.
As for you talking about how smart you think you are…lol. Whatever helps you hold on to the identity your actual life isn’t confirming I guess.
And, he’s back,
Implicit in the definition is doing well on tests relative to others not themselves. If we know their performances relative to one another on two tests we have no reason to think that the next score of an individual will further regress, in my words, much if at all.
Swank, the example is a typical lawyer at an average law school who did super well on the LSAT, and we want to predict how he’ll score on the WAIS-IV. Now you’re saying we know other test scores beyond the LSAT. How do we know how well the typical lawyer who did well on the LSAT did on previous tests? We don’t, but we can predict his previous scores from his LSAT but then we’re using his predicted previous scores to predict his WAIS-IV score so it’s a prediction of a prediction which is wrong and redundant because we’re still just using his LSAT to predict his WAIS-IV, we’re just doing it indirectly.
But we do. He was part of a pre-defined group who took the SAT. So if we adjust to an average SAT of 1200 with an SD of around .6*200 ~ 120, we can calculate going forward. It’d be an LSAT of something like 160 with an sd of 6 rather than 10.
We have those scores, we can estimate. Then we can restrict the range again within the same law school.
Long story short, in this better approximation the difference is much more than 12 IQ points. The law school SD is 4-6 LSAT points.
And the separate issue regarding law school selecting independently for IQ, when you reduce it to pass/fail….not so much. You may as well select individuals to go on a long hike based on LSAT.
So yes I know what you’re doing and I’ve known the whole time. There are better things to do.
But we do. He was part of a pre-defined group who took the SAT. So if we adjust to an average SAT of 1200 with an SD of around .6*200 ~ 120, we can calculate going forward. It’d be an LSAT of something like 160 with an sd of 6 rather than 10.
We have those scores, we can estimate. Then we can restrict the range again within the same law school.
Swank, we’re trying to estimate the WAIS-IV IQ of the typical super high LSAT lawyer from a typical law school. Thus by definition, all we know is he got super high LSATs and attended a typical law school.
Now you’re saying we know more because such a person probably had high SAT scores. But “probably had high SAT scores” is already implied by the fact that he’s both a lawyer and had super high LSAT scores so using “probably high SAT scores” as third indicator of his likely WAIS-IV scores is redundant. It provides no independent information about the individual; it’s just a deduction based on information already provided.
In other words, you don’t understand how regression works because you don’t understand the difference between a dependent variable and an independent variable.
It’s not probably high SAT scores. It’s he HAS a high SAT score relative to other lawyers.
It’s an independent variable because in the real world every future lawyer in fact HAS an SAT score relative to future lawyers. A future lawyer in fact has an LSAT score relative to future lawyers . He in fact has two scores that are easy to find out without estimation.
When you say “we’ re trying to…” you keep confusing me saying “we can do this instead of that” with “if we do that then this will happen.”
In other words you still don’t understand.
‘Thus by definition,’
And just in case you don’t think I’ve been saying ‘you don’t need to use this definition’ in so many many words:
I have said there are better things to do. I have said that you do not need to do that, etc. etc. so many times, yet each time you come back to me, you come back to me with the assumption of ‘if we define it this way, then…’
So it is definitively you who does not understand.
It’s not probably high SAT scores. It’s he HAS a high SAT score relative to other lawyers.
It’s an independent variable because in the real world every future lawyer in fact HAS an SAT score relative to future lawyers. A future lawyer in fact has an LSAT score relative to future lawyers . He in fact has two scores that are easy to find out without estimation.
That doesn’t change anything, Swank. Lawyers from a typical law school have SAT scores that are predictable from the distribution of LSAT scores of lawyers from a typical law school, given the correlations of SATs with occupation, law school and LSAT scores independent of one another . So even if you get the actual SAT scores, they don’t provide any new information beyond what is already predictable from the information we have.
When you say “we’ re trying to…” you keep confusing me saying “we can do this instead of that” with “if we do that then this will happen.”
I keep confusing it because it’s functionally the same thing.
I have said there are better things to do.
There are always better things to do Swank. You can imagine scenarios where we have more data than we do, though it’s just predictable from the data we already have.
So it is definitively you who does not understand.
You’re right; I don’t understand gibberish.
‘So even if you get the actual SAT scores, they don’t provide any new information beyond what is already predictable from the information we have.’
If you define the group before the tests, which would mean adjusting the means and SDs on each test beforehand, and you have the actual SAT and LSAT score, you simply will not expect the future score to regress. If John takes three IQ tests, on test 3 he won’t be regressing to the mean of whatever group he was in on test 1, which is this case would be ‘future lawyers.’ Instead, we would expect the scores on the second retest to match the scores on the first retest.
And this is why restricting the range to ‘within law school’ would be the only restriction that would diminish the LSAT score’s accuracy.
If all we know about John is he’s a future lawyer with a score of A on the SAT and a score of B on the LSAT & attended law school C, then John’s score on the LSAT would be expected to regress on the WAIS-IV to whatever is the mean WAIS-IV score of future lawyers who scored A on the SAT and attended law school C.
If John is typical of future lawyers who scored B on the LSAT and attended law school C, then his score on the SAT (even if we know it) could be predicted by regressing from his score on the LSAT to the mean SAT of future lawyers at law school C. If John is not typical, then he not a particularly revealing case study, so either way, knowing his SAT doesn’t really tell us anything new about the IQ gap between high LSAT lawyers and low LSAT lawyers from the same law school .
Which is why I said the only limitation on the LSAT’s accuracy for the IQ of a typical lawyer would be the within-school restriction. I’ve said it several times. However, if we’re just talking about a typical lawyer’s LSAT score, the fact that lawyers have been heavily selected for ‘it’s not luck driving your score,’ e.g. undergraduate tiers of future lawyers aligning almost perfectly with law school tiers, plus the fact that by and large high-LSAT-relative-to-lawyers are also high-SAT-relative-to-lawyers and we have the scores to show this….means that a lawyer with a high LSAT is not likely regressing to the mean of lawyers.
In sum, this statement ‘[b]ecause of a statistical phenomenon known as range restriction, among lawyers, especially lawyers at the same law school, the LSAT probably only correlates about 0.3 ‘ is horseshit.
Which is why I said the only limitation on the LSAT’s accuracy for the IQ of a typical lawyer would be the within-school restriction. I’ve said it several times.
As usual your statements are vague and ambiguous so God knows what you mean this time, but if you mean that the only reason why the LSAT correlation with the WAIS-IV would diminish among lawyers is because they attend the same law school, then no, even among all American lawyers, the correlation between the LSAT and the WAIS-IV should be lower than it is in the entire LSAT taking population which itself would be lower than the correlation obtained if all law school age Americans took both the LSAT and the WAIS-IV
However, if we’re just talking about a typical lawyer’s LSAT score, the fact that lawyers have been heavily selected for ‘it’s not luck driving your score
Luck is relative Swank. If a random person scores 125, that’s probably partly luck, but when a lawyer scores IQ 125 on a test, then’s probably not luck because that’s typical of lawyers.
But when a lawyer scores 150, that’s so high it shows luck, even if you’re a lawyer.
.,’ e.g. undergraduate tiers of future lawyers aligning almost perfectly with law school tiers,
It’s not almost perfect, and attending an elite ungrad school helps you get accepted into an elite law school, even holding grades and test scores constant, so it’s not strong evidence of people scoring high repeatedly.
In sum, this statement ‘[b]ecause of a statistical phenomenon known as range restriction, among lawyers, especially lawyers at the same law school, the LSAT probably only correlates about 0.3 ‘ is horseshit.
As I’ve acknowledged from the outset, 0.3 might be an underestimate; the point is the correlation would not be high.
Once again, it wouldn’t diminish the LSAT’s accuracy when we have redefined the group, ie already adjusted the SD and mean, pre SAT.
Yes luck is relative and typical lawyers who have scored above the lawyer mean twice have not gotten luckier than the typical lawyer. They were selected for that non-randomness. They are not regressing to that mean.
Going to a university in the same tier as one’s desired law school does not, on its own, contribute much.
Once again, it wouldn’t diminish the LSAT’s accuracy when we have redefined the group, ie already adjusted the SD and mean, pre SAT
Are you arguing it wouldn’t be diminished, or that it wouldn’t be further diminished? Two very different arguments..
Yes luck is relative and typical lawyers who have scored above the lawyer mean twice have not gotten luckier than the typical lawyer. They were selected for that non-randomness. They are not regressing to that mean.
People regress to the mean of the group they belong to. The more good things we know about someone (i.e. did well on the SAT, passed the bar, etc) the higher the mean we expect them to regress to. In my example, all we knew was that John was a typical super high LSAT lawyer from a typical law school. Of course, from that example you can deduce a lot more, but deductions by definition are not independent information.
Going to a university in the same tier as one’s desired law school does not, on its own, contribute much.
I disagree. If you apply to law school as a Harvard undergrad, you can get away with having lower grades and lower LSAT scores than an undergrad from some nothing college.
If I have said diminish with the qualifier redefined the group, overall I must mean further diminish since we defined the group in a way that restricted range from outset.
Having a high LSAT score relative to other lawyers informs you enough about an individual, especially a typical lawyer, that the score will not likely change much, if at all…..unless you further restrict the range.
The restriction you’re talking about wouldn’t be that major.
I’ve seen you say that many times about Ivy undergrads but whatever the add is probably amounts to one point on the LSAT. It is way overstated. If one wants to remain in the Ivy one must continue to perform at that level.
Having a high LSAT score relative to other lawyers informs you enough about an individual, especially a typical lawyer, that the score will not likely change much, if at all…..unless you further restrict the range.
Regression occurs anytime there is an imperfect correlation between two tests among the people who take them because all regression is, is the slope of the line of best fit through a scatter plot showing the scores on both tests expressed as Z scores relative to the people being tested. So unless you think the LSAT and WAIS-IV correlate virtually perfectly among lawyers, substantial regression to the lawyer mean should occur when predicting the scores of lawyers from one test to the other.
I’ve seen you say that many times about Ivy undergrads but whatever the add is probably amounts to one point on the LSAT. It is way overstated. If one wants to remain in the Ivy one must continue to perform at that level.
So if two lawyers apply to Yale law school…both are identical in every way (same GPA, same race, same letters of recommendation, same SES, same region of the country, etc) except one is from Harvard and scored 170 on the LSAT and the other is from a very low SAT college and scored 180 on the LSAT. Who will Yale Law pick? There’s a good chance they’ll reject the higher LSAT person because they’ll think, there’s obviously something wrong with him if he’s so smart yet still ended up at a crappy undergrad college. We don’t want freaks.
Yes, duh, pumpkin. The typical lawyer (versus the typical LSAT taker) with a high LSAT score has two separate high scores relative to other lawyers. The typical lawyer also remained within whatever tier he was in for undergraduate school. An atypical lawyer may have split his scores widely and dramatically moved tiers, but a typical lawyer’s scores are pretty much what they are, which is why the correlation between his LSAT and WAIS will be very high, ergo he will not regress much, if at all, to the mean of lawyers.
‘Who will Yale Law pick? There’s a good chance they’ll reject the higher LSAT person’
A) they wouldn’t reject him based on ‘we don’t want freaks,’ and B) in this instance they would assuredly take the higher LSAT score —- provided the higher LSAT scorer’s grades are somewhat competitive.
If two individuals with LSAT 170 apply to Yale, and one of these individuals went to an Ivy…sure, then the Ivy individual may get in over the public-U 170. But 180 versus 170? It did not happen.
Yes, duh, pumpkin. The typical lawyer (versus the typical LSAT taker) with a high LSAT score has two separate high scores relative to other lawyers. The typical lawyer also remained within whatever tier he was in for undergraduate school. An atypical lawyer may have split his scores widely and dramatically moved tiers, but a typical lawyer’s scores are pretty much what they are, which is why the correlation between his LSAT and WAIS will be very high, ergo he will not regress much, if at all, to the mean of lawyers.
So your argument is that among a random sample of all American lawyers of a given age (not lawyers at the same law school), the correlation between LSAT and scores on some other respected IQ test (i.e. WAIS-IV, SAT) would be virtually perfect (0.9+)?
And your evidence for this is that many students who attended Ivy League law schools also have Ivy League undergrad degrees, and thus you believe they scored very similarly on the SAT and the LSAT? Is that the argument?
‘the correlation between LSAT and scores on some other respected IQ test (i.e. WAIS-IV, SAT) would be virtually perfect (0.9+)?’
It’d be very high. SAT and LSAT scores are already highly correlated around .9. Granted, restricting the range on the SAT would diminish the correlation, but because we have two similar scores to work with, this would matter less. They are two tests, and they are two tests of different skills — the LSAT and SAT-V have some overlap, though. Some of the LSAT’s ‘e,’ in T+e involves that portion of IQ that the LSAT doesn’t test. The SAT does test much of that, which means that knowing an individual’s SAT score would increase the LSAT’s accuracy, because we would know that it wasn’t just ‘luckily not being tested on X type of IQ,’ in the ‘e’ portion.
‘And your evidence for this is that many students who attended Ivy League law schools also have Ivy League undergrad degrees, and thus you believe they scored very similarly on the SAT and the LSAT? Is that the argument?’
Evidence of what? That Ivy League law school attendance is driven primarily by LSAT score and that Ivy League undergrad status doesn’t matter, or that typical lawyers have LSAT scores that are very similar to their SAT scores?
If you’re referring to the latter, then the observation is that a typical lawyer does not change school prestige tiers between SAT and LSAT.
I think you’re overestimating the correlation between the SAT & LSAT in even an unrestricted sample, underestimating the decline in correlation caused by range restriction, overestimating the correlation between law school prestige and undergrad prestige, and overestimating the role of scoring high on both tests in causing that correlation.
‘I think you’re overestimating the correlation between the SAT & LSAT in even an unrestricted sample’
The data for SAT score by college major versus LSAT score by college major suggests otherwise.
‘underestimating the decline in correlation caused by range restriction,’
Doubtful. The LSAT correlates highly with IQ tests, and it correlates with ‘g’ to the same extent as most IQ tests afaik. A .81 correlation would reduce to about .6 if we restricted the range to ‘lawyers.’ A further range restriction (within same law school) would push it down to about ~.55.
The correlation would then increase with knowledge of the above-group-average SAT score, because some amount of variance in the ‘e’ term arises from ‘luck resulting from measuring a certain subsection of IQ.’ That ‘e’ term diminishes with knowledge of the SAT and gives us more confidence that the LSAT mark is the true score. If it explains even just 10-20 percent of the additional variance, the correlation shoots to .68-.75 from .6.
‘overestimating the correlation between law school prestige and undergrad prestige,’
Go look at the LSAT scores by undergraduate college and add a few points (the pool is future lawyers, not undergraduates that took the LSAT), and you will see that there really isn’t much mobility.
‘and overestimating the role of scoring high on both tests in causing that correlation.’
Even though prestige of law school varies almost perfectly with LSAT and doesn’t show much of a relationship with GPA (or anything else for that matter…the reason why there seems to be a relationship is because attending an undergrad Ivy greatly increases the odds that someone will have an ‘Ivy worthy’ score).
Doubtful. The LSAT correlates highly with IQ tests, and it correlates with ‘g’ to the same extent as most IQ tests afaik. A .81 correlation would reduce to about .6 if we restricted the range to ‘lawyers.’ A further range restriction (within same law school) would push it down to about ~.55.
The correlation would then increase with knowledge of the above-group-average SAT score, because some amount of variance in the ‘e’ term arises from ‘luck resulting from measuring a certain subsection of IQ.’ That ‘e’ term diminishes with knowledge of the SAT and gives us more confidence that the LSAT mark is the true score. If it explains even just 10-20 percent of the additional variance, the correlation shoots to .68-.75 from .6.
This is very strange logic. The SAT scores of high LSAT lawyers would only be high to the extent that the LSAT correlates with other IQ tests including the SAT among lawyers (from a particular law school). So if you believe the correlation between the LSAT and other tests diminishes, then knowledge of high SAT scores can’t save the correlation, because the SAT scores themselves would be dragged down by the diminished correlation.
Now if you believe the SAT correlates much higher with LSATs than other IQ tests do, then they’re essentially so similar that the LSAT provides no new information.
Either way, it’s gibberish.
Go look at the LSAT scores by undergraduate college and add a few points (the pool is future lawyers, not undergraduates that took the LSAT), and you will see that there really isn’t much mobility.
Based on the SATs used to select them, Harvard undergrads have an IQ of 141. Based on their LSAT scores, they have an IQ of maybe 132. 32/41 = 0.78 correlation between SATs and LSATs in the general population since I’m dividing distances from the U.S. mean of 100. Now an IQ of 141 is probably 28 points higher than the LSAT population, and an IQ of 132 is probably 19 points higher than the LSAT population. 19/28 = 0.68 correlation between LSATs and SATs in the LSAT population. Expect the correlation to be even lower in the lawyer population and even lower still in the lawyer from same law school population, and expect comparable dropping correlations between the LSAT and the WAIS-IV.
The SAT can’t save the correlation between the LSAT and the WAIS-IV anymore than the WAIS-IV can save the correlation between the LSAT and the SAT.
‘
The SAT and LSAT would both be “dragged down” but because they measure different things, knowledge of one would increase the accuracy of the other.
A typical lawyer has similar scores pumpkin, it’s foolhardy to believe otherwise for all the other reasons I’ve mentioned.
Harvard undergrads and Harvard LSAT takers aren’t really typical. The correlations between the majors are high. So it’s likely somewhere between .78-.9. Maybe you think that’s low.
And like I said, it doesn’t have to give us that much more information to substantially change the correlation.
Add now the fact that these are multiple observations and that the individuals have been selected for a certain level of performance twice. They simply did not get that lucky and the ‘e’ is much lower.
There really is no way around the fact that at minimum your original figure is at most half the likely figure, first of all, and second of all, that having multiple, different tests of any individual will increase the expected accuracy of the score.
There really is no way around the fact that at minimum your original figure is at most half the likely figure,
I agreed from the outset that the 0.3 figure was just a guess. It depends on how high the correlation between the LSAT and WAIS-IV is in the general population and how much range restriction there is among lawyers from the same law school.
first of all, and second of all, that having multiple, different tests of any individual will increase the expected accuracy of the score.
There is no second of all Swank. Get that idea out of your head. Performance on previous tests is already factored in by regressing lawyers to the mean of their particular law school. In my example I chose an average law school, but if you want to imagine an elite law school, you would just regress to a higher mean.
In the typical case, all valid adult IQ tests a typical lawyer at a given law school ever took (whether at school, on the internet, in a psychologist’s office) will be normally distributed around the score predicted by regressing his LSAT score to the mean IQ of his classmates on any other IQ test. Thus arguing that these other test scores, predictable from his LSAT, add new information is just a figment of your imagination.
It would be true, in the individual case, but not in the typical case.
You made a plainly unreasonable guess from the outset.
There is a second of all. To make it super simple, let’s just say that among typical lawyers, the LSAT correlates more highly with, IQ tests than it would for normal individuals in the general population.
And it seems like you’re getting your 141 IQ number from the SAT I conversion chart. If we want to assume the empirically verified .81 correlation, then 1450-900/220 *.81 = 130 IQ. The LSAT IQ score is almost 1-1.
Oh when you said LSAT IQ, you didn’t mean LSAT-to-IQ, apparently. Lol. That’s even sillier. John (or Ted, I can’t remember at this point)’s 180 LSAT-to-IQ would not = IQ 152…it’d be something like 145. The 150 LSAT would still be around IQ 112 or so. Regardless, the IQ distance between them would remain high.
Restricting the range doesn’t change much, either. Mean IQ 125, SD 8-10 versus 160 LSAT SD 6. 20/6*.6*9 +125 = 143 IQ for 180 LSAT and 114 IQ for 150 LSAT among future lawyers.
Not much difference…as I said.
And it seems like you’re getting your 141 IQ number from the SAT I conversion chart. If we want to assume the empirically verified .81 correlation, then 1450-900/220 *.81 = 130 IQ. The LSAT IQ score is almost 1-1.
Sigh. Swank Harvard students are a subset of the general U.S. population selected largely for especially high SAT scores. The fact that multiplying their deviation from the mean by 0.81 reduces them to 30 IQ points above 100, means that pre-multiplied, they were roughly 140. In other words, you’re agreeing with me that even in a general population sample, the correlation between tests is far from perfect. Now just imagine how low the correlation gets among lawyers from the exact same law school, where range restriction is extreme..
Restricting the range doesn’t change much, either. Mean IQ 125, SD 8-10 versus 160 LSAT SD 6. 20/6*.6*9 +125 = 143 IQ for 180 LSAT and 114 IQ for 150 LSAT among future lawyers.
And this is exactly what I did in the original post, except I guessed a correlation (in the restricted groups of lawyers from a typical law school) of 0.3, while you’re assuming 0.6. The true figure is probably somewhere in between.
I’m not agreeing with you. When you say LSAT/SAT IQ will significantly regress to the mean, a normal person reading assumes that you mean that the IQ derived from their LSAT or SAT will ‘significantly’ regress to the mean. What you actually mean to say is that their SAT or LSAT scores relative to their peers won’t be as high as their IQ scores relative to their peers.
The reason why ‘we multiply’ is that the term ‘SAT IQ’ is meaningless unless you are actually deriving an IQ from an SAT score — which you weren’t. You were just assuming their SAT scores relative to their peers doubled as ‘SAT IQ,’ which is ‘strange logic.’
Of course deriving an IQ from an SAT score relative to one’s peers still would leave one expecting some sort of ‘regression.’ However, knowing that the SAT-derived-IQ and the LSAT-derived-IQ are nearly the same, tells us that the correlations between tests are high enough for us not to expect much, if any regression.
You keep asserting (based on nothing) that within-law-school range restriction is extreme. How many times do I have to tell you that the range restriction won’t lessen the correlation much below .6 and that the .6 correlation won’t really make much of a difference for their IQ-derived-from-LSAT? Infinitely, apparently. Sigh, indeed.
In sum, the guess was facially unreasonable, and any assertion that we could expect IQs derived from LSATs to show significant regression is false.
I’m not agreeing with you. When you say LSAT/SAT IQ will significantly regress to the mean, a normal person reading assumes that you mean that the IQ derived from their LSAT or SAT will ‘significantly’ regress to the mean. What you actually mean to say is that their SAT or LSAT scores relative to their peers won’t be as high as their IQ scores relative to their peers.
When I say LSAT IQ of SAT IQ, I simply mean LSAT scores or SAT scores converted to the IQ scale where the general U.S. population has a mean of 100 and an SD of 15. That is, if every American adult in their 20s took the LSAT, whatever their median score was would be assigned an LSAT IQ of 100 and whatever the top 2% scored on the LSAT would be assigned an IQ of 130. That’s all anyone means when they do they SAT/LSAT IQ conversions. It’s not something I invented Swank. They’re all over the web.
The reason why ‘we multiply’ is that the term ‘SAT IQ’ is meaningless unless you are actually deriving an IQ from an SAT score — which you weren’t.
See above.
Of course deriving an IQ from an SAT score relative to one’s peers still would leave one expecting some sort of ‘regression.’ However, knowing that the SAT-derived-IQ and the LSAT-derived-IQ are nearly the same, tells us that the correlations between tests are high enough for us not to expect much, if any regression.
When a group of people is selected based on independent criteria, the IQs are the same. When they are selected for one, they regress on the other.
You keep asserting (based on nothing) that within-law-school range restriction is extreme.
Not based on nothing. Look at the LSAT difference between the 25 percentile and the 75 percentile at a typical law school. The difference is small.
How many times do I have to tell you that the range restriction won’t lessen the correlation much below .6 and that the .6 correlation won’t really make much of a difference for their IQ-derived-from-LSAT? Infinitely, apparently. Sigh, indeed.
And how many times do I have to tell you that 0.3 was just a guess and the true figure could be twice that. In other words, I conceded that point from the outset. But then you argued over and over that there would be little or no regression. How could there possibly be no regression? A correlation of 0.6 implies 40% regression, assuming the correlation is even that high.
In sum, the guess was facially unreasonable, and any assertion that we could expect IQs derived from LSATs to show significant regression is false.
Only if you consider over 40% insignificant
‘That’s all anyone means when they do they SAT/LSAT IQ conversions.’
No it isn’t. All the conversions I have seen take into account a) the test-taking population’s mean IQ, b) the test-taking population’s IQ SD, and c) the correlation between the test and an actual IQ score. I have never seen anyone do what you said they do, so I doubt that method is ‘all anyone means.’ An LSAT score is not an IQ score and there is no such thing as ‘LSAT IQ’ in the sense you are talking about. LSAT IQ is shorthand for ‘IQ derived from LSAT,’ which involves knowing (or taking a guess) at the test-taking population’s mean and SD and converting using the correlation.
MENSA and the high IQ societies most assuredly don’t ‘mean’ that. A qualifying MENSA test score for the SAT pre-1994 was 1250 — definitely not the top 2% of test-takers, and the LSAT is the same.
‘Look at the LSAT difference between the 25 percentile and the 75 percentile at a typical law school. The difference is small.’
The typical law school difference is around 3 points from 50th-75th. From 50th to 84th, it’d be around 4-5 points. The SD of typical lawyers on the LSAT would be ~6 points.
‘How could there possibly be no regression? A correlation of 0.6 implies 40% regression, assuming the correlation is even that high.’
The only ‘regression’ would come from the fact that the LSAT is not an IQ test. LSAT scores are not IQ scores. Like I said, LSAT IQ is meaningless if you aren’t actually converting the score. When you convert the scores and get an actual ‘LSAT IQ,’ the range restriction makes little difference, because the range restriction comes with smaller SD’s and a higher mean. Take LSAT 170. 20/10*.81*13.2 + 113 = 135. 10/6*.6*9 + 125 = 134. Nearly the same score, pumpkin.
‘And how many times do I have to tell you that 0.3 was just a guess’
An unreasonable guess. The entire post suffers greatly from your bias against Ivy leaguers and in favor of IQ having a great effect on income.
No it isn’t. All the conversions I have seen take into account a) the test-taking population’s mean IQ,
Right
b) the test-taking population’s IQ SD,
Right
and c) the correlation between the test and an actual IQ score.
Wrong.
I have never seen anyone do what you said they do, so I doubt that method is ‘all anyone means.’ An LSAT score is not an IQ score and there is no such thing as ‘LSAT IQ’ in the sense you are talking about. LSAT IQ is shorthand for ‘IQ derived from LSAT,’ which involves knowing (or taking a guess) at the test-taking population’s mean and SD and converting using the correlation.
No wonder you were arguing there would be no more regression. You thought the scores I was using were already regressed. You seem to think that when people convert college admission test scores into IQ, they are predicting what someone with that SAT/LSAT score would get on an IQ test. They are not.
Instead, they are assuming the SAT and LSAT are disguised IQ tests, and are simply changing the scale to the IQ scale.
MENSA and the high IQ societies most assuredly don’t ‘mean’ that. A qualifying MENSA test score for the SAT pre-1994 was 1250 — definitely not the top 2% of test-takers, and the LSAT is the same.
But they are assuming that the test takers are a select group, and that if every young adult in America took those tests, only the top 2% would score 1250 pre-1994.
Like I said, MENSA’s numbers only make sense with a correlation, so c) is likely true.
‘You seem to think that when people convert college admission test scores into IQ, they are predicting what someone with that SAT/LSAT score would get on an IQ test. They are not.’
Such a belief isn’t even necessary for what I am saying. But no, they don’t just 1-1 test SD and test-taking IQ SD. It did not happen.
‘But they are assuming that the test takers are a select group, and that if every young adult in America took those tests, only the top 2% would score 1250 pre-1994.’
They are multiplying the SAT score by a correlation, pumpkin. That’s it. That is why post-1994 they said that SAT score no longer sufficiently correlate with an IQ test: they are very much aware of the correlation with actual IQ. (Keep in mind post-1994 it still correlates >.8 as well….so the correlation must have been higher according to them). They could have said ‘the 2% level has changed’ or whatever, but they didn’t.
Like I said, MENSA’s numbers only make sense with a correlation, so c) is likely true.
I don’t think so. For example when Bush’s 1206 old SAT score were converted to IQ, it sounds like it was done so by finding what percentage of the entire U.S. population (his age) could score higher on the SAT, and then assigning that percentile the corresponding IQ score:
Linda Gottfredson, co-director of the University of Delaware-Johns Hopkins Project for the Study of Intelligence and Society, told United Press International: “I recently converted Bush’s SAT score to an IQ using the high school norms available for his age cohort. Educational Testing Service happened to have done a study of representative high school students within a year or so of when he took the test. I derived an IQ of 125, which is the 95th percentile.” In other words, only one out of 20 people would score higher.
Another IQ expert, Charles Murray of the American Enterprise Institute, the co-author of the bestseller “The Bell Curve,” came up with a similar result when asked by UPI. Noting that everybody except high school dropouts takes the PSAT when they are sophomores, Murray calculated from PSAT scores that “I think you’re safe in saying that Dubya’s IQ, based on his SAT score, is in excess of 120, which puts him the top 10 percent of the distribution, but I wouldn’t try to be more precise than that.”
http://www.upi.com/Business_News/Security-Industry/2004/01/14/Analysis-How-smart-is-Bush/39471074106535/
’Further, the Prometheus Society (which accepts anyone with an IQ above 160-one in 30,000 level) accepts people who scored above the one in 30,000 level on the old SAT (not on some theoretical IQ test predicted from the SAT):
The difference between 1600 and 1560 is typically 2 to 4 problems on the “old” (pre-recentered) SAT. However, when figuring percentile equivalents for the SAT, it should be remembered that it is based upon a sample size of approximately 1 million actual test takers selectively sampled from a general population size in excess of 3 million. It isn’t unreasonable to assume that the general population percentiles that we assign to the SAT at the top end (for which selection is the highest) are accurate for the test group as a whole. In fact, however, in a population of 3 million there should be over 100 individuals scoring at the 1-in-30,000 level. On any given year less than ten individuals obtained a perfect score on the old SAT with on the order of 100 or less scoring 1560 or more and, therefore, it is is safe to say that the 1-in-30,000 level is achieved by these individuals.
http://216.224.180.96/~prom/oldsite/mcreport/memb_comm_rept.html#8.5.6 Establishing a credible 1-in-30,000 of the
Further, if IQs of 160 (for Prometheus society) were regression predictions from SAT scores, then the actual SAT scores would have to be in the top several million level. Not even a perfect old SAT score was that rare.
They are multiplying the SAT score by a correlation, pumpkin. That’s it. That is why post-1994 they said that SAT score no longer sufficiently correlate with an IQ test: they are very much aware of the correlation with actual IQ. (Keep in mind post-1994 it still correlates >.8 as well….so the correlation must have been higher according to them). They could have said ‘the 2% level has changed’ or whatever, but they didn’t.
In order to justify using the SAT as an IQ test, they have to show it correlates with other IQ tests. That’s why they mention the correlation. The same thing is done with official IQ tests. Correlations with other official IQ tests are cited as a way of establishing validity. But your score is your score on the test you take, not a statistical prediction of how you would have scored on some other test.
The Bush IQ estimation is explained by insufficient data from a certain time period.
The Prometheus IQ justification is explained mostly by them accounting for the SAT’s somewhat low ceiling. Even 1-1 SD conversions don’t match up.
They might use r^2 to derive the IQ score and use SAT sd*r as their IQ sd too.
I’m not saying people never use percentile matching, but it’s certainly not all they use.
Regarding MENSA and validity….I don’t think that’s why. Post-1994 the correlation remains high.
Regardless it doesn’t seem like you’re ever going to concede that the guess was unreasonable.
The Bush IQ estimation is explained by insufficient data from a certain time period.
No it’s explained by the fact that they consider the SAT an IQ test, and thus are converting his SAT percentile directly to the IQ score with that percentile.
The Prometheus IQ justification is explained mostly by them accounting for the SAT’s somewhat low ceiling. Even 1-1 SD conversions don’t match up.
But a 1-1 percentile match up does.
I’m not saying people never use percentile matching, but it’s certainly not all they use.
The only example I’m aware of, where someone converted SATs/LSATs to IQ using a regression prediction was a paper by Frey and Detterman, and it resulted in such low IQ predictions, that it did not become popular.
Regarding MENSA and validity….I don’t think that’s why. Post-1994 the correlation remains high.
MENSA probably abandoned the new SAT out of pressure from the college board, that didn’t want SATs associated with a topic as controversial as IQ. But the point is the 1250 old SAT Mensa cutoff was based on the assumption that if all American 17 year olds took the SAT, only 2% would have scored 1250+
It was NOT based on the likely score of 1250 SAT score people on an official IQ test.
Regardless it doesn’t seem like you’re ever going to concede that the guess was unreasonable.
I conceded that a long time ago Swank.
Way back in January I wrote in response to you:
0.3 was just a guestimate, but not an especially far-fetched one. The correlation between Raven IQ and SAT scores was only around 0.4 before correction for range restriction, and lawyers from the same law school are more range restricted than college undergrads at the same college
On Feb 1 I wrote,
I don’t know what the exact correlation between IQ and LSAT scores is among lawyers who attended the same law school; all I know is that it’s a heck of a lot lower than 0.81. 0.3 was just a guess based on what little data I could find.
On Feb 3, I wrote:
As I said from the beginning, 0.3 was just an educated guess on my part but I haven’t thoroughly investigated it; the true figure could be more than double that. But the point is, the individual lawyers at a specific school would regress to the mean of their school on a second testing…by how much depends on the correlation between the two tests in that restricted population
So you had a valid point about 0.3 being low, and I conceded it.
So why didn’t this discussion end back on Feb 3 or even January? Because you had this strange idea that because the typical high LSAT lawyer has taken multiple tests, his scores would not regress when he takes yet another.
Now in your defense, maybe it’s because you thought I was using scores that were already regressed since you seem to think that’s how college admission tests are normally converted into IQ equivalents.
So let’s just leave it at that, otherwise we’ll still be discussing this in March.
None of what you wrote before suggests the guess was unreasonable. If that’s what you meant, then point taken.
Identical degrees no there, only similar degrees (seems identical), because two ”phd” in same specialization for example, are official, technic and uniform result about different individuals, with different intelligence combinations, with different paths, different motivations.
Extrinsical or intrinsical Motivation is a very important piece to be analysed.
Your question is trivial. Obviously there a correlation among iq, income and similar degrees, lol… 😉
Always will be a correlation, because we are all correlated one each other, everything is correlated…
Look at to university teachers. They seemed ”have” similar ‘iq’, income and degrees… In stable professions, this correlation will be positive, in unstable professions, i.e, the correlation will be weak, because will be more competition and different results by different levels of personal motivation.
Therefore, the answer to your question is: YEAAAAHHH….
Pingback: Brain size and IQ correlated with family income - Playa del Carmen, Mexico forum