Razib Khan One-stop-shopping for all of my content

June 24, 2012

Higher vocabulary ~ higher income

Filed under: data,Data Analysis,GSS,Income,IQ — Razib Khan @ 7:54 pm

Prompted by a comment below I was curious as to the correlation between intelligence and income. To indicate intelligence I used the GSS’s WORDSUM variable, which has a ~0.70 correlation with IQ. For income, I used REALINC, which is indexed to 1986 values (so it is inflation adjusted) and aggregates the household income. Finally, I limited my sample to non-Hispanic whites over the age of 30 (for what it’s worth, this choice also limited the data set to respondents from the year 2000 and later).

The results don’t get at the commenter’s assertions, because 10 out of 10 on WORDSUM does not imply that you’re that smart really. But the trendline is suggestive. Note that aggregated 0-4 because the sample size at the lower values is small indeed.

April 15, 2012

Common variant for “IQ gene”?

Filed under: Genetics,Genomics,Human Genetics,Human Genomics,IQ — Razib Khan @ 10:11 pm

A few people have forwarded me this paper, Identification of common variants associated with human hippocampal and intracranial volumes:

…Whereas many brain imaging phenotypes are highly heritable…identifying and replicating genetic influences has been difficult, as small effects and the high costs of magnetic resonance imaging (MRI) have led to underpowered studies. Here we report genome-wide association meta-analyses and replication for mean bilateral hippocampal, total brain and intracranial volumes from a large multinational consortium. The intergenic variant rs7294919 was associated with hippocampal volume (12q24.22; N = 21,151; P = 6.70 × 10−16) and the expression levels of the positional candidate gene TESC in brain tissue. Additionally, rs10784502, located within HMGA2, was associated with intracranial volume (12q14.3; N = 15,782; P = 1.12 × 10−12). We also identified a suggestive association with total brain volume at rs10494373 within DDR2 (1q23.3; N = 6,500; P = 5.81 × 10−7).

Look at the sample sizes. Beware of behavior genomics with small sample sizes. Paul Thompson, one of the many authors of this paper, is giving media interviews. To me that’s a good sign, as he’s a very smart guy. He has some confidence in this study. Here’s the section which is resulting in ...

January 1, 2012

Too smart to be a good cop

Filed under: IQ,Psychology — Razib Khan @ 3:35 pm

Several readers have pointed me to this amusing story, Court OKs Barring High IQs for Cops:

A man whose bid to become a police officer was rejected after he scored too high on an intelligence test has lost an appeal in his federal lawsuit against the city.

“This kind of puts an official face on discrimination in America against people of a certain class,” Jordan said today from his Waterford home. “I maintain you have no more control over your basic intelligence than your eye color or your gender or anything else.”

Jordan, a 49-year-old college graduate, took the exam in 1996 and scored 33 points, the equivalent of an IQ of 125. But New London police interviewed only candidates who scored 20 to 27, on the theory that those who scored too high could get bored with police work and leave soon after undergoing costly training.

The average score nationally for police officers is 21 to 22, the equivalent of an IQ of 104, or just a little above average.

But the U.S. District Court found that New London had “shown a rational basis for the policy.” In a ruling dated Aug. 23, the 2nd Circuit agreed. The court said the policy might be unwise but was a rational way to reduce job turnover.


First, is the theory empirically justified? If so, I can see where civil authorities are coming from. That being said, it’s obvious that there are some areas where “rational discrimination” is socially acceptable, and others where it is not. The same arguments used to be applied to women, in terms of the actuarial probabilities that they would get pregnant and so have to leave the workforce. And disparate impact always looms large in the utilization of these sorts of tests.

Second, can’t you just fake a lower score on an intelligence test? Do police departments hire statisticians to smoke out evidence of conscious selection of incorrect scores? I doubt it. Jordan may be smart, but perhaps he lacks common sense if the upper bound for IQ was well known.

My initial thought was that an IQ of 104 seemed too low for a median police officer, but poking around it does seem plausible as a descriptive statistic. Honestly I don’t have much acquaintance with the police, so I’ll trust the scholars no this. That being said, is it in our social interest for police officers to be so average? I don’t know. Though is it in the social interest that someone with an IQ as high as Robert Jordan’s ends up a prison guard?

June 27, 2010

Psychometrics, epigenetics and economics

Filed under: IQ,Psychology,Psychometrics — Razib Khan @ 10:26 am

Two papers of interest. IQ in the Production Function: Evidence from Immigrant Earnings (ungated). And Human Intelligence and Polymorphisms in the DNA Methyltransferase Genes Involved in Epigenetic Marking. My impression is that the focus on epigenetics has a higher-order social motive; even the sort of humanists who are involved with N + 1 have asked me about the topic. But how many people know what methylation is?

May 4, 2010

WORDSUM & IQ & the correlation

Filed under: Blog,data,Data Analysis,GSS,IQ,WORDSUM — Razib Khan @ 2:08 pm

Every time I use the WORDSUM variable from the GSS people will complain that a score on a 10-question vocabulary test is not a good measure of intelligence. The reality is that “good” is too imprecise a term. The correlation between adult IQ and WORDSUM = 0.71. The source for this number is a 1980 paper, The Enduring Effects of Education on Verbal Skills. I’ve reproduced the relevant table…

Estimated Correlations for Variables in a Model of Enduring Effects of Education for White, Native-Born People 25 to 72 Years Old in the Contemporary [1970s] United States
  Child IQ Age Sex Father’s Educ Father’s SEI Educ Adult IQ WORDSUM
Child IQ - 0 0 0.31 0.30 0.51 0.80 -
Age - - 0.026 -0.304 -0.130 -0.304 -0.42 -0.005
Sex - - - -0.054 0.058 0.050 0 -0.121
Father’s Educ - - - - 0.488 0.469 0.30 0.302
Father’s SEI - - - - - 0.347 0.31 0.285
Educ - - - - - - 0.66 0.511
Adult IQ - - - - - - - 0.71
WORDSUM - - - - - -   -
                 

Obviously since the WORDSUM test was not given to those under 18 you can’t calculate the correlation between childhood IQ and WORDSUM score. Additionally, I suspect since 1980 there’s been a bit more cognitive stratification by education. I notice in the GSS sample that there are many older people, especially women, who have high WORDSUM scores but no college education. In the younger age cohorts this pattern is not as evident because if you are intelligent the probability is much higher that you’ll obtain a university education.

A correlation of 0.71 is not mind-blowing, there’s a significant difference between IQ and WORDSUM as they relate to each other linearly. But I think it’s good enough to get a sense that WORDSUM is a serviceable substitute for a more rigorous measure of g in lieu of any alternatives, and not so clumsy a proxy so as to be useless. Though that call is up to you, and readers are free to disagree with the methodology of the model used to obtain this correlation. Additionally, I would point out that WORDSUM is a subset of the vocabulary subsection of the Wechsler Adult Intelligence Scale. WORDSUM is in effect a slice of an IQ test.

I am bookmarking this post so that in the future I can simply place a link in the comment threads in response to objections to WORDSUM.

Note: Thanks to Bryan Caplan for pointing me to this paper.

Citation: Lee M. Wolfle, Sociology of Education, Vol. 53, No. 2 (Apr., 1980), pp. 104-114

February 1, 2010

Half Sigma’s flawed post on DTNBP1

Filed under: Behavior Genetics,IQ — ben g @ 10:33 am

A while back, Mark and I were working on a comprehensive post which would try to tally the results of the various IQ-gene studies to see what they said about racial differences. We began this quest bright-eyed and hopeful that we would help contribute to ending a calamitous debate that has gone on for way too long. However, as we learned more about genetics, and these studies in particular, we came to realize that it’s too early to take IQ-genes seriously.

We began with an approach similar to what Half Sigma did 2 years ago with the DTNBP1 gene. However, we soon learned that this approach was incredibly flawed and misleading. I wasn’t going to write this post, but recently Half Sigma’s DTBP1 post was linked from Reddit and tens of thousands of people are viewing it. When I saw that, I frustratedly criticized HS. He responded that I should give a more diplomatic and reasoned response, so here it is:

  1. You cannot simply add up SNPs from the same gene or chromosome. Half Sigma simply adds the observed effects of the SNPs to one another, ignoring that the alleles are highly correlated with one another, and not independently inherited, which is referred to as linkage disequilibrium (LD). The study that Half Sigma used provides the following table of LD for its SNPs:

    rs2619539 rs3213207 rs1011313 rs2619528 rs760761 rs2619522 rs2619538
    rs2619539 0.156 0.111 0.0 0.0 0.001 0.055
    rs3213207 1.0 0.014 0.334 0.403 0.34 0.076
    rs1011313 0.916 1.0 0.037 0.033 0.036 0.081
    rs2619528 0.024 0.955 1.0 0.838 0.737 0.128
    rs760761 0.015 1 1.0 0.96 0.854 0.166
    rs2619522 0.04 0.955 1.0 0.867 0.96 0.182
    rs2619538 0.242 0.823 0.825 0.648 0.772 0.778

    As can be seen in this table, pairwise LD goes as high as 1.0, meaning that two of the alleles are always inherited together. Adding these SNP’s together is therefore like counting them twice.

  2. Group comparisons require replication in both groups. Because different populations have systematic genetic and environmental differences, an effect in one group may not occur in another. The study that Half Sigma uses relies primarily on a (small) sample of Dutch people. It is unclear whether these effects would exist in a population of African ancestry, let alone another European one.
  3. Candidate-gene association studies are not reliable. This is the most important point. Candidate gene association studies have largely failed to replicate. In fact, there have been no common IQ polymorphisms which have been replicated. Genome-wide association studies, which don’t suffer as severely the various biases of candidate-gene association studies like publication bias or the winner’s curse have not shown common SNP-associations with IQ.

    IQ is highly heritable, so the problem is the current methods, not the search for genes. With the development of sequencing technology and huge cohorts, we will be able to see the genes that are really behind normal IQ variation. With replication in multiple ethnicities and races, we will also see to what extent various genes and environments are responsible for group differences. There’s no need to make proclamations of victory for hereditarianism or environmentalism in the mean time.

December 8, 2009

Does the family matter for adult IQ?

Filed under: Behavior Genetics,IQ — ben g @ 5:57 am

A frequent claim in the IQ debates is that which family you are raised in has no lasting impact on your IQ. Jensen argues in The g Factor that the only causes of IQ similarities between adult identical twins are genetic. Many researchers go so far as to argue that by 12 years of age, the shared environment has no impact.

Based on my limited knowledge of the behavior genetic research, I used to hold this position as well. But thanks to some recent in depth reading, I have come to the conclusion that which family you are raised in matters significantly for your IQ as an adult, especially so for people of lower socioeconomic status. I’ll detail the behavior genetic evidence here, and argue that it points to significant shared environmental influences on adult IQ scores.

Twin Studies

The most recent and comprehensive survey of twin studies on IQ comes from Haworth et al (2009). Using pooled twin data from around the world, they modeled genetic and environmental influences as a function of age. Here is what they found regarding the effects of the shared environment:

[S]hared environment shows a decrease from childhood (33%) to adolescence (18%) but remained at that modest level in young adulthood (16%).

In an email exchange with Dr. McGue (one of the co-authors of the paper) he told me that while the latest data may not fit with earlier estimates, it’s actually more reliable due to the unprecedented sample size (11,000 pairs of twins).

One failing of this study, though, is that it doesn’t go far enough into adulthood. The young adult group ranges from 14 to 34 years of age, with an average age of 17. In contrast, McGue (1993) looked seperately at data on adults over 20 years of age. He found that the shared environment diminished to zero impact at that point. Here’s his chart:
Looking at that chart, you might quickly conclude that shared environmental influence evaporates by age 20. However, this conclusion is premature. Twin studies make a great number of assumptions, some of which increase and others of which decrease estimates of the shared environment. A straightforward way of bypassing these assumptions is to compare monozygotic twins reared apart (MZAs) to monozygotic twins reared together (MZTs). The following data comes from a comparison of MZTs and MZAs, of average age 41, in Bouchard (1990):

Measure MZA correlation MZT correlation
WAIS IQ-Full Scale 0.69 0.88
WAIS IQ-Verbal 0.64 0.88
WAIS IQ-Performance 0.71 0.79

Differences between MZA’s and MZT’s on Raven’s Progressive Matrices follow the same pattern but are even more extreme. Bouchard (1981) reported a median correlation of only 0.58 for adult MZA’s on the Raven’s. Curiously, though, MZA’s are equally if not more correlated than MZT’s on the Mill-Hill vocabulary test. Apparently, the pattern is that more g-loaded tests tend to show stronger evidence of lasting shared environmental impact.

It’s worth noting that MZT vs. MZA comparisons are actually biased towards an underestimation of shared environmental impact. Bouchard’s study of twins reared apart found an environmental correlation of .22 for MZAs on various environmental measures, with some having a small but significant correlation with IQ scores. Also MZA’s share the womb. To summarize: when the assumptions of the twin method are effectively controlled for, lasting shared environmental impacts are revealed.

Adoption Studies

To date, most adoption studies of IQ have concluded that being adopted by a new and typically well-off family has no effect on adult IQ scores. Here is a chart of adoption studies from Bouchard (2009):As you can see by clicking it, the IQ correlation between unrelated individuals in the same family decreases (on average) from .26 in childhood to .04 in adulthood (which begins at age 17 for the purposes of this graph).

However, as with the previous chart, the quick conclusion that shared environmental influences don’t matter in adulthood shouldn’t be so quickly accepted. To begin with, we can see that the adoption data underestimates the shared environment relative to the twin literature. This most likely occurs because of the assumptions that go into adoption studies.

Stoolmiller (1999), for example, highlighted the issue of range restriction– the idea that the limited range of adoptee and adoptive family environments will lower estimates of the shared environment. This idea is supported by studies which make the extra effort to include individuals of lower SES. The French adoption studies that made such an effort buck the trendline seen above, in finding that nurture matters almost as much as nature for the IQ of 14 year olds. Scarr (1993) is the outlier in the adoption graph above, finding a .19 correlation between unrelated adolescent siblings. Perhaps her results differed from others because her sample was multi-racial and therefore less range restricted. Lastly, there are other lines of evidence supporting the idea of range restriction, such as Turkheimer’s work on SES and cognitive ability.

It’s worth noting, however, that McGue (2007) looked for evidence of range restriction effects within the “broad middle class” and did not find any. He used statistical methods that are over my head to estimate the effects of range restriction based on a range restricted sample and state census data. Unfortunately there are no studies which have critiqued his as of yet. Any commenters who are familiar with the statistics involved are invited to comment. Even if McGue is right about restriction of range, my point stands that assumptions inherent in the adoption studies deflate c^2 estimates.

Future Directions

Future work will help sort out the still unanswered question of shared environmental influences on adult IQ scores. There are large longitudinal adoption studies currently under way, and I believe that Haworth’s twin study will be followed-up on and include data on older twins. There are also interesting (albeit less methodologically agreed upon) studies coming out like this one, which find significant shared effects on IQ in adulthood.

My reading of the available evidence is that there is a significant shared environmental input to adult IQ, and that it is associated with socioeconomic status. To what extent it’s the neighborhood or the parents themselves that matters is unclear. Just as the most g-loaded tests show the most shared environmental effects in the MZA-MZT comparison, so too does the Flynn effect occur on the most g-loaded tests, suggesting that whatever is loading onto the “shared environment” within generations is also responsible for differences between them.

Does the family matter for adult IQ?

Filed under: Behavior Genetics,IQ — ben g @ 5:57 am

A frequent claim in the IQ debates is that which family you are raised in has no lasting impact on your IQ. Jensen argues in The g Factor that the only causes of IQ similarities between adult identical twins are genetic. Many researchers go so far as to argue that by 12 years of age, the shared environment has no impact.

Based on my limited knowledge of the behavior genetic research, I used to hold this position as well. But thanks to some recent in depth reading, I have come to the conclusion that which family you are raised in matters significantly for your IQ as an adult, especially so for people of lower socioeconomic status. I’ll detail the behavior genetic evidence here, and argue that it points to significant shared environmental influences on adult IQ scores.

Twin Studies

The most recent and comprehensive survey of twin studies on IQ comes from Haworth et al (2009). Using pooled twin data from around the world, they modeled genetic and environmental influences as a function of age. Here is what they found regarding the effects of the shared environment:

[S]hared environment shows a decrease from childhood (33%) to adolescence (18%) but remained at that modest level in young adulthood (16%).

In an email exchange with Dr. McGue (one of the co-authors of the paper) he told me that while the latest data may not fit with earlier estimates, it’s actually more reliable due to the unprecedented sample size (11,000 pairs of twins).

One failing of this study, though, is that it doesn’t go far enough into adulthood. The young adult group ranges from 14 to 34 years of age, with an average age of 17. In contrast, McGue (1993) looked seperately at data on adults over 20 years of age. He found that the shared environment diminished to zero impact at that point. Here’s his chart:
Looking at that chart, you might quickly conclude that shared environmental influence evaporates by age 20. However, this conclusion is premature. Twin studies make a great number of assumptions, some of which increase and others of which decrease estimates of the shared environment. A straightforward way of bypassing these assumptions is to compare monozygotic twins reared apart (MZAs) to monozygotic twins reared together (MZTs). The following data comes from a comparison of MZTs and MZAs, of average age 41, in Bouchard (1990):

Measure MZA correlation MZT correlation
WAIS IQ-Full Scale 0.69 0.88
WAIS IQ-Verbal 0.64 0.88
WAIS IQ-Performance 0.71 0.79

Differences between MZA’s and MZT’s on Raven’s Progressive Matrices follow the same pattern but are even more extreme. Bouchard (1981) reported a median correlation of only 0.58 for adult MZA’s on the Raven’s. Curiously, though, MZA’s are equally if not more correlated than MZT’s on the Mill-Hill vocabulary test. Apparently, the pattern is that more g-loaded tests tend to show stronger evidence of lasting shared environmental impact.

It’s worth noting that MZT vs. MZA comparisons are actually biased towards an underestimation of shared environmental impact. Bouchard’s study of twins reared apart found an environmental correlation of .22 for MZAs on various environmental measures, with some having a small but significant correlation with IQ scores. To summarize: when the assumptions of the twin method are effectively controlled for, lasting shared environmental impacts are revealed.

Adoption Studies

To date, most adoption studies of IQ have concluded that being adopted by a new and typically well-off family has no effect on adult IQ scores. Here is a chart of adoption studies from Bouchard (2009):As you can see by clicking it, the IQ correlation between unrelated individuals in the same family decreases (on average) from .26 in childhood to .04 in adulthood (which begins at age 17 for the purposes of this graph).

However, as with the previous chart, the quick conclusion that shared environmental influences don’t matter in adulthood shouldn’t be so quickly accepted. To begin with, we can see that the adoption data underestimates the shared environment relative to the twin literature. This most likely occurs because of the assumptions that go into adoption studies.

Stoolmiller (1999), for example, highlighted the issue of range restriction– the idea that the limited range of adoptee and adoptive family environments will lower estimates of the shared environment. This idea is supported by studies which make the extra effort to include individuals of lower SES. The French adoption studies that made such an effort buck the trendline seen above, in finding that nurture matters almost as much as nature for the IQ of 14 year olds. Scarr (1993) is the outlier in the adoption graph above, finding a .19 correlation between unrelated adolescent siblings. Perhaps her results differed from others because her sample was multi-racial and therefore less range restricted. Lastly, there are other lines of evidence supporting the idea of range restriction, such as Turkheimer’s work on SES and cognitive ability.

It’s worth noting, however, that McGue (2007) looked for evidence of range restriction effects within the “broad middle class” and did not find any. He used statistical methods that are over my head to estimate the effects of range restriction based on a range restricted sample and state census data. Unfortunately there are no studies which have critiqued his as of yet. Any commenters who are familiar with the statistics involved are invited to comment. Even if McGue is right about restriction of range, my point stands that assumptions inherent in the adoption studies deflate c^2 estimates.

Future Directions

Future work will help sort out the still unanswered question of shared environmental influences on adult IQ scores. There are large longitudinal adoption studies currently under way, and I believe that Haworth’s twin study will be followed-up on and include data on older twins. There are also interesting (albeit less methodologically agreed upon) studies coming out like this one, which find significant shared effects on IQ in adulthood.

My reading of the available evidence is that there is a significant shared environmental input to adult IQ, and that it is associated with socioeconomic status. To what extent it’s the neighborhood or the parents themselves that matters is unclear. Just as the most g-loaded tests show the most shared environmental effects in the MZA-MZT comparison, so too does the Flynn effect occur on the most g-loaded tests, suggesting that whatever is loading onto the “shared environment” within generations is also responsible for differences between them.

November 21, 2009

Models of IQ & wealth

Filed under: IQ — Razib @ 5:48 pm

Steve Hsu has been interesting of late (interesting like Steve, not Malcolm). So, IQ, compression and simple models and If you’re so smart, why aren’t you rich?. For a theoretical physicist I find Steve to be eminently clear in his exposition of abstract topics (perhaps he has practice from having to talk to experimental physicists?).

November 8, 2009

The quest for common variants & cognition

Filed under: Association,IQ,Population genetics — Razib @ 11:39 am

A genome-wide study of common SNPs and CNVs in cognitive performance in the CANTAB:

Psychiatric disorders such as schizophrenia are commonly accompanied by cognitive impairments that are treatment resistant and crucial to functional outcome. There has been great interest in studying cognitive measures as endophenotypes for psychiatric disorders, with the hope that their genetic basis will be clearer. To investigate this, we performed a genome-wide association study involving 11 cognitive phenotypes from the Cambridge Neuropsychological Test Automated Battery. We showed these measures to be heritable by comparing the correlation in 100 monozygotic and 100 dizygotic twin pairs. The full battery was tested in 750 subjects, and for spatial and verbal recognition memory, we investigated a further 500 individuals to search for smaller genetic effects. We were unable to find any genome-wide significant associations with either SNPs or common copy number variants. Nor could we formally replicate any polymorphism that has been previously associated with cognition, although we found a weak signal of lower than expected P-values for variants in a set of 10 candidate genes. We additionally investigated SNPs in genomic loci that have been shown to harbor rare variants that associate with neuropsychiatric disorders, to see if they showed any suggestion of association when considered as a separate set. Only NRXN1 showed evidence of significant association with cognition. These results suggest that common genetic variation does not strongly influence cognition in healthy subjects and that cognitive measures do not represent a more tractable genetic trait than clinical endpoints such as schizophrenia. We discuss a possible role for rare variation in cognitive genomics.

David Goldstein is one of the authors. I wonder if this influenced his views on the evolution of intelligence.

Powered by WordPress