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

July 25, 2018

The Insight show notes: episode 30, Genetics and educational attainment

Filed under: Education,Genetics,Psychology — Razib Khan @ 3:47 pm

This week Razib and Spencer discussed the relationship between educational attainment and genetics on The Insight (Apple Podcasts, Stitcher and Google Play) with James Lee, lead author of Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals (published in Nature Genetics).

Here are some more resources: FAQs about “Gene discovery and polygenic prediction from a 1.1-million-person GWAS of educational attainment”. The Atlantic and The New York Times also covered the paper. An op-ed in The New York Times, Why Progressives Should Embrace the Genetics of Education.

The three laws of behavior genetics and the fourth law of behavior genetics are both mentioned. The study was a meta-analysis of genome-wide associations (GWAS), and may have been the largest GWAS published to date.

Much of the discussion centered around intelligence. The podcast with Stuart Ritchie was cited as a useful primer (remember to subscribe with Apple Podcasts, Stitcher and Google Play). You might want to check out Ritchie’s book, Intelligence.

Population stratification was mentioned. Martin et al., and two preprints, Berg et al. and and Sohail et al., tackle this issue in relation to disease and height, and how it confounds our understanding. Lee discussed LD score regression as a way to account for stratification in this particular analysis..

There was extensive discussion of the concept of heritability, where genetics explains variation in a trait.

The Social Science Genetic Association Consortium (SSGAC) and its research projects were referenced extensively.

Each allele seems to effect ~1 week of education. The authors returned more than 1,000 statistically significant markers.

Spencer brought up the “omnigenetic” model. This comes from Boyle et al., An Expanded View of Complex Traits: From Polygenic to Omnigenic.

James mentioned some of Camille Benbow’s work, in particular Life Paths and Accomplishments of Mathematically Precocious Males and Females Four Decades Later.

Interested in learning where your ancestors came from? Check out Regional Ancestry by Insitome to discover various regional migration stories and more!


The Insight show notes: episode 30, Genetics and educational attainment was originally published in Insitome on Medium, where people are continuing the conversation by highlighting and responding to this story.

July 23, 2018

The genetics of education

Filed under: Education,Genetics,Intelligence,Psychology — Razib Khan @ 11:49 am
Yale University

In the modern world, obtaining an education is a rite of passage. Not only does education provide one with skills useful for the modern economy, but it also helps to form one’s values and socializes one with peers who go through the same life experiences. Education isn’t just learning about various disciplines, it is a way to learn how to live in the modern world.

It is a topic which intersects with sociology, politics, and even ethics. As it turns out, education, or the attainment thereof, also intersects with genetics. This follows naturally from the first law of behavior genetics, “all human behavioral traits are heritable.” By “heritable,” geneticists refer to the fact that variation of a trait correlates with variation in genes. That variation tracks a causal relationship — so that genetic variants in some way cause a particular outcome.

This is easy enough to illustrate with an example. Imagine a genetic variant that changes the production of a biochemical that impacts whether someone is hyperactive or not. Hyperactivity is a behavioral characteristic with a lot of variables. Someone who drinks too much coffee will exhibit hyperactivity. But, it is surely true that some of the variation on these personality traits are due to cognitive neurological differences — some of which are then due to genetic differences between people. We may not be “born that way,” but we are probably “disposed to be that way.”

That’s a lot of caveats, and that is accounted for in the third law of behavior genetics: “a substantial portion of the variation in complex human behavioral traits is not accounted for by the effects of genes or families.” When it comes to behavior, environment matters. Although, what constitutes “environment” is not always clear, but any understanding of the genetic basis of behavioral variation needs to account for the fact that much of behavioral variation has nothing to do with genetics.

Galton’s classic illustration of parent-child correlation on height

And a when it comes to educational attainment, there is obviously no one “gene for education.” Whether or not you obtain a degree is impacted by many factors: from family encouragement and resources, inspirational teachers, intelligence, and your own conscientiousness. But, some of these characteristics, in particular the ones having to do with intelligence and personality, are impacted by your genes.

It has long been known through indirect methods that intelligence and personality are heritable. Identical twins are much more similar on these characteristics than conventional siblings, and relatives are much more similar than non-relatives. But, finding the biophysical genetic basis has been difficult because of the fourth law of behavior genetics: “A typical human behavioral trait is associated with very many genetic variants, each of which accounts for a very small percentage of the behavioral variability.”

A traditional way for geneticists to discover the cause of a trait is to target particular genetic locations and see if they are associated with the trait in question. This “candidate gene” method has been useful for many diseases, where a single defective mutation is responsible for much of the cause of the disease. But, it has been an utter failure in behavior genetics because of the fourth law of behavior genetics. To establish a connection between a genetic variant and a behavioral trait requires enormous sample sizes and a good knowledge of the human genome.

Until the year 2000, we didn’t have the sequence of a human genome, and until the past decade, dense assays of human genomic variation were expensive — this meant that studies were limited to small sample sizes and only a few genes. Most of the published results did not replicate, because they were not true in terms of the effect of the gene on the trait in question.

Recently, all that has changed. Thanks to “next generation sequencing” and “chip technology” researchers now have access to hundreds of thousands of markers in any given person — and cheaply at that. This cost-effectiveness allows for an increase in sample size; as many more people can be tested. This shatters the barriers implied by both the third and fourth laws of behavior genetics: small effect sizes no longer impede discovering ‘the needles in the haystack’ of complex traits. Bigger sample sizes and more subtle statistical methods are producing results that only a few years ago would have seemed fantastical.

A new paper in Nature Genetics illustrates this starkly, Gene discovery and polygenic prediction from a genome-wide association study of educational
attainment in 1.1 million individuals
. The authors identified 1,271 independent genome-wide-significant SNPs. This is a big achievement, considering that five years ago a paper with ~125,000 individuals identified just 3 SNPs that were significant for this trait!

Though it is hard to generalize about 1,000+ genetic variants, the figure to the left illustrates that the genes that these variants are found in are highly over-expressed in the nervous system. This is exactly what you see in most genetic analyses of complex traits that are behavioral. The genetic “hits” are found disproportionately in genes that control variation in neurological function because behavior is downstream of brain function. To be fair, many genes express in the brain, so that’s not a surprise. Rather, the authors compared the gene’s expression level to the typical gene.

Curiously these hits are not particularly over-expressed in genes associated with the development of glial cells, those cells in the central nervous system which are not neurons. Because these cells form the tissue which scaffolds the connections between neurons, the authors suggest that this might mean that differences in cognitive ability between individuals may not be a function of “transmission speed.” This highlights the fact that the these sorts of abstruse statistical analyses ultimately aim to uncover underlying biological phenomena.

And yet such a paper, with over 1 million samples from numerous cohorts, will have to get into the statistical weeds. One of the major issues that crops up in these analyses is “stratification.” This means that the genetic variation in the sample is correlated with variables such as geographical population structure. Therefore, some of the positive hits for any of these sorts of analyses might easily be picking up the overall population genetic variation and differences between groups, which may not have a genetic basis at all (e.g., British tend to drink tea, Americans tend to drink coffee).

Empirical genetic relationship of siblings

To get around this, the authors look at a sub-sample of 20,000 sibling-pairs. Many of the issues presented by population stratification do not apply within families. Families have the same broad genetic background, and also control for many environmental differences (since siblings are raised in the same family and socioeconomic context). But, there is still genetic variation among siblings, and some of this variation is responsible for variation in traits between siblings. After all, height tends to run in families, but the difference in height between same-sex siblings is not usually due to differences in nutrition (at least in the developed world).

Looking at the associations between genetic variation and educational attainment within families the authors found “that within-family
effect sizes are roughly 40% smaller than GWAS effect sizes.” In other words, there are factors that seem to result in the overestimation of the genetic effects on educational attainment within the broader population. The authors note that the same does not apply to height.

What might account for this then? One possibility is that some of the genes that a parent has, but does not transmit to the offspring, might result in a more beneficial environment. This is often termed the “parental effect.”

The paper looked are more than just educational attainment. With sample sizes in the hundreds of thousands they also looked at cognitive performance and mathematical ability (self-reported). Using the same methods as for educational attainment, the authors predicted around 10% of the variance.

Of course there are limitations. The sample size is large, but not diverse genetically. Overwhelmingly of European origin, the authors found that their method could predict less than 5% of the variance in African Americans. This is not surprising, because genome-wide associations often do not predict well across different populations.

Additionally, there is the reality that these methods focus on common variation within populations. The heritability of most behavioral traits using more indirect classical methods is much higher than this ~10% of variation explained would imply, so there is still a genetic component to be accounted for. Perhaps this variation is found in rare genetic variants, which are not explored in this sort of research.

Ultimately, we may look back at this 1 million-person analysis as the first in a scholarly tradition of massive GWAS sample sizes. Genomics is cheap enough that it is possible that genetic sample sizes in the range of a billion are feasible within 15 years. That will probably require a whole different set of esoteric methods but will probably yield many novel results.

Interested in learning where your ancestors came from? Check out Regional Ancestry by Insitome to discover various regional migration stories and more!


The genetics of education was originally published in Insitome on Medium, where people are continuing the conversation by highlighting and responding to this story.

April 11, 2018

The “g” in genes

Filed under: behavioral-science,Genetics,Psychology,science — Razib Khan @ 1:12 pm
What’s next?

Intelligence, or smarts, is once of those words which has many meanings. That’s why we say “street smart” or “book smart.” When psychologists speak of intelligence, however, they are usually referring to something more precise and specific. The image above is a sample of a question from the Raven’s Progressive Matrices test, which is “used in measuring abstract reasoning and regarded as a good non-verbal estimate of fluid intelligence.”

Fluid intelligence “is the capacity to reason and solve novel problems, independent of any knowledge from the past.”

When I was an undergraduate student, my physics professor would often assign problems on the exam which had no explicit corollaries in the problem sets or lecture. During one after-exam review session, one student brought this issue up, and the professor simply offered that given what we learned in the course, we should be able to “derive a method to solve a general class of problem.” I suspect from the distribution of scores that more than half the time a typical student couldn’t derive a method in the allotted period. I know this was often the case with me.

In relation to tests which measure one’s analytic skills or cognitive tasks like memory recall, researchers have found that outcomes are positively correlated. If you do well on one test of this sort, you tend to do well on another such test.

The variable which summarizes these correlations is termed the “general intelligence factor,” often just shortened to g.

When it comes to intelligence, this is what psychologists are really interested in — not the outcome on one specific test. General intelligence is the most distilled and reduced aspect of “book smarts” that psychologists have been able to construct.

So what good is it? More than half of the variation in academic achievement is predicted by variation in g. People in higher status and higher paying jobs tend to have higher general intelligence. And higher g also correlates with a longer lifespan. Because of these correlations it is no surprise that intelligence testing was originally used to identify children who were not performing as well as their peers, and see if they might benefit from special attention.

Not only does general intelligence correlate with many things in one’s life, there is also a correlation between parents and offspring. The most recent work suggests that about 50% of the variation in general intelligence in the population can be accounted for by variation of genes. That is, intelligence is 50% heritable.

Multivariate Gaussian distribution

The implication here is that though parents and children, or siblings, may exhibit a correlation, it is imperfect. The brilliant mathematician Carl Friedrich Gauss came from unexceptional parents, and his numerous descendants are not particularly exceptional. In contrast, the Bernoulli family were a literal mathematical dynasty.

For a complex trait which exhibits a distribution, there are many variables at play, and genes are just one of them. Because so many genes seem to control behavioral and cognitive traits, such as intelligence, until recently, we couldn’t pinpoint any specific region of the genome which impacted variation on these characteristics within the normal range.

With modern genomic methods, which survey variation across the whole genome across huge numbers of people, this is changing. For example, a new paper establishes links to variation in intelligence at over 500 genes! This is still a small number in the grand scheme of things, but whereas five years ago we didn’t know any genes associated with intelligence, today we know hundreds.

A “chip” which asseses thousands of genetic markers

Though indirect methods, such as comparing correlations with and across families, allow us to arrive at a 50% proportion for what is heritable in intelligence, known genomic variation only accounts for a few percent of this heritable component as of this writing. But within the year, it seems likely that the 10% value barrier will be broken, and eventually we may know most of the genetic positions that account for the heritable component of intelligence within human populations.

Then the full story can begin to be told, because once we start to establish the boundaries of the genetic basis of intelligence, we can explore the environmental territory — which accounts for the other 50% our intelligence.

Explore your Regional Ancestry story today.


The “g” in genes was originally published in Insitome on Medium, where people are continuing the conversation by highlighting and responding to this story.

January 21, 2018

The rich can afford to not steal

Filed under: Psychology,Replication crisis — Razib Khan @ 12:07 pm

Pardis Sabeti has an op-ed in The Boston Globe, For better science, call off the revolutionaries. In it, she contrasts and compares the changes in social psychology and genetics over the past 15 years. In the former, you have had the replication crisis, while in the latter you saw the confirmation that many candidate-gene studies were not robust and replicable.

Sabeti observes that while there has been a great deal of vehemence in the change in social psychology, in genetics people have taken the overturning of the old studies in stride, and not personalized it.

I think we can all agree that personal attacks, vehemence, and Jacobin behavior is counter-productive to what science aims to be. This applies outside of the context of the replication crisis in social psychology. The problem emerges when you ask what the alternative strategy is to overturn an entrenched order.

Though it’s implied in the piece, the reason that geneticists could be more graceful probably has more to do with the situation they were presented with in the 2000s. The rise of genomics and computation generated a surfeit of data and the ability to analyze that data. Repeated gains in power and precision mean that every overturned orthodoxy gave rise to many more research opportunities. An established researcher whose candidate-gene study was overturned had bigger fish to fry than defend turf which was small potatoes set next to the possibilities of the “post-genomic” future.*

In other words, the difference here is situational. I was aware there was a problem in social psychology by the middle-2000s, because I had friends and acquaintances who were graduate students in the field. They complained about all the things that we now know were problems. It was one of those “open secrets” where less powerful people couldn’t speak the truth, and more powerful people who benefited from the status quo had no incentive to change the norms. I really don’t know how this was going to change in a gentle fashion.

* Compare the arguments between selectionists and neutralists in the 1970s vs. now. The reality is most people just want to analyze the data, not argue about theoretical issues. That’s because we have data.

December 19, 2017

Motivated reasoning in “science journalism.”

Filed under: Psychology,Psychometrics — Razib Khan @ 6:09 pm

The “reproducibility crisis” has really benefited some sectors of science journalism, as there is less credulous amplification of spurious results. That being said, motivated reasoning is powerful. They “want to believe.”

So when I saw this piece in Quartz, Highly motivated kids have a greater advantage in life than kids with a high IQ, I immediately scanned for what I usually look for, and found it:

Over the next four decades, the Gottfrieds and several colleagues collected a staggering trove of data on the study participants, yielding important insights into working parents, temperament, and other topics. Researchers collected information about participants from parents, teachers and transcripts, tested their IQ and motivation levels,and even visited their homes. In all, the Fullerton Longitudinal Study has amassed an estimated 18,000 pieces of information on each of the remaining 107 participants. “It’s our life’s work,” says Allen cheerfully. “We’ll take it to our grave.”

107 participants. Lots of information huh? Things that make you go hm….. Also, 19% of the children had IQs of 130 or above. About 2% of the population has an IQ at this level. The sample size was relatively small, and the sample was very unrepresentative.

This doesn’t mean that there aren’t real results in these data. But I don’t think they warrant the fanfare in the title, except for the fact that people want a silver bullet that will abolish social inequality.

Even the text itself doesn’t justify the title at all (to be fair, usually headline writers differ from the persons writing the text of a piece): “[Motivation] in itself is accounting for a certain amount of variance in achievement that goes above and beyond IQ….” That is, they don’t even say it accounts for more of the variance, only that there is variance that isn’t accounted for by IQ (which everyone already agreed upon).

Finally, I’ve spent my life around highly educated and intelligence people a bit perplexed and befuddled by my diverse interests. This includes in academia. So I can see that there is difference between people for whom learning is a means to a professional and social ends, and for those whom learning is the ends. I suspect the ancients could have told you this!

October 22, 2017

The backlash against social psychology was pent up demand

Filed under: Psychology,Social Psychology — Razib Khan @ 12:02 pm

Both Slate and When the Revolution Came for Amy Cuddy, have pieces deconstructing the fall from grace of an idea like “power posing.”

This is all obviously wrapped up in the “replication crisis”, which is impacting most sciences which use some statistics and are generally characterized by modest and complex causal effects (social and biological sciences in particular then).

Obviously, I am no social psychologist, but can I just say that everyone knew there was a problem in the field. By everyone, I mean psychologists. I had friends who worked in related fields who told me as early as 2006 not to trust anything coming out of social psychology. Others described how p-hacking and “unconscious” data manipulation was relatively common in psychological experimentation, and the personal stands they had to take.

When everyone knows that something is wrong, but no one says anything, you have a coordination problem. But once the snowball starts rolling down the hill…everyone decides to speak their mind.

Finally, there’s the demand-side problem: ideas like power posing, implicit bias, and stereotype threat, offer neat, clean, and powerful explanations and oftentimes solutions for social problems. Wonkish Left-liberal publications and pundits in particular literally mine the literature to “show what the science says” (don’t worry, it overwhelmingly confirms prior beliefs).

As a testament to power of the likely wrong (not robust) viewpoints, consider that John Bargh has a book out, Before You Know It: The Unconscious Reasons We Do What We Do. Bargh’s work was one of the first research programs to be critiqued in the early 2010s. Of course he doesn’t agree with the critics, but it does strike me that the field as a whole (e.g., people like Daniel Kahneman) believe that these subliminal effects are much weaker than originally claimed, at best. Nevertheless Bargh is going to sell his books, and people in coffee shops and airports all over the country are going to eat it up.

April 13, 2017

The coming reign of the Baby Boomer gerontocracy

Filed under: Gerontocracy,Psychology,Psychometrics — Razib Khan @ 1:22 pm

From Dawn to Decadence: 1500 to the Present: 500 Years of Western Cultural Life is one of my favorite books. It’s one of those works whose breadth and depth is such that I would recommend it to anyone. Jacques Barzun began writing this work when he was 84, and it was published in his 93rd year. Born in 1907 Barzun saw the full efflorescence of 20th century Western culture across much of its span firsthand. When people say that when you age you gain wisdom, surely in the domain of scholarship Barzun’s production in the last few decades of his life would qualify.

But not everyone is Jacques Barzun. If you read Intelligence: All That Matters or peruse some of Eliott Tucker-Drob’s work you will know that cognitive function declines with age beyond your twenties. Different subcomponents may decline at different rates. And, they decline differently in different people (e.g., some people may develop dementia, so their faculties will decline far faster at an earlier age). But, by and large any gains in experience or wisdom are going to be balanced against declines in raw analytic ability, as well as the slow entropic loss of information.

This is not an inconsequential matter. Our governing class is quite old. The average age in Congress may be 55 to 60, but it is almost certainly true that more senior members with more power and authority are older. The president of the United States is 70 years old. If you look at the plots in these figures by 70 there has been a notable drop in intelligence by this age, though again, it may vary from person to person.

But most important in light of these figures is that the Supreme Court is a lifetime appointment, and many of its members are quite old, an anticipate serving until they are quite old if they are younger. In the mid-1970s justice William O. Douglas had a stroke and was basically not mentally competent to serve. Because of this fact, and Douglas’ reluctance to retire his fellow justices basically did not take his vote into account. Three of the justices today are over the age of 70, with Clarence Thomas nearing that age, and two are over the age of 80.

When it comes to Congress, or even the President, there seems to be some sort of institutional support as well as the larger collective vote in the case of Congress, which might buffer the cognitive impact of a gerontocracy. But aside from law clerks Supreme Court justices have to rely on their own individual mental capacities.

The Mormon Church has a gerontocracy among its we openleadership. Even my most devout friends in the church sometimes found it amusing how old their leadership was, and how quickly they died in succession due to the seniority principle. But The Supreme Court is not the leadership of a relatively small church. It impacts our whole nation. This sort of gerontocracy is no laughing matter.

Will we openly speak of the age issue? I doubt it. Today the Baby Boomers are between the ages of 53 an 71. They are coming into their own as a cohort into the highest reaches of the gerontocracy. If there is any generation with the grace and humility to step aside for the greater good, it will not be this generation.

October 30, 2012

Which results from cognitive psychology are robust & real?

Filed under: Cognitive Psychology,Psychology — Razib Khan @ 8:11 pm

A paper on the psychology of religious belief, Paranormal and Religious Believers Are More Prone to Illusory Face Perception than Skeptics and Non-believers, came onto my radar recently. I used to talk a lot about the theory of religious cognitive psychology years ago, but the interest kind of faded when it seemed that empirical results were relatively thin in relation to the system building (Ara Norenzayan’s work being an exception to this generality). The theory is rather straightforward: religious belief is a naturally evoked consequence of the general architecture of our minds. For example, gods are simply extensions of persons, and make natural sense in light of our tendency to anthromorphize the world around us (this may have had evolutionary benefit, in that false positives for detection of other agents was far less costly than false negatives; think an ambush by a rival clan).*

 

But enough theory. Are religious people cognitively different from those who are atheists? I suspect so. I speak as someone who never ever really believed in God, despite being inculcated in religious ideas from childhood. By the time I was seven years of age I realized that I was an atheist, and that my prior “beliefs” ...

September 7, 2012

Who believes in I.Q.?

Filed under: Psychology — Razib Khan @ 12:49 am

There are many things that a given individual believes which are ‘heterodox’ in their social circle. For example, I have long thought that intelligence tests are predictive of life outcomes, and somewhat heritable in a genetic sense (these are both true, the objection of skeptics usually rests on the fact that they are skeptical of the construct itself). As I have explained here before I did not always hold to these views. Rather, when I was in seventh grade a teacher who mentored me somewhat took me aside after class, and suggested that perhaps some of my slower classmates were not quite as lazy as I obviously presumed (I tended to get impatient during mandatory group projects). When I was 5 years old and starting kindergarten my command of English was rather weak, and my mother explained to me that Americans were a very smart people. By the end of the year I was excelling. Throughout my elementary school years I frankly had a smugness about me, because I accepted what my parents told me, that academic outcome is a function of the virtue of effort. And I had quite a bit of virtue if the results were any gauge.

But ...

July 19, 2012

Cousin marriage can reduce I.Q. a lot

Filed under: I.Q.,Psychology — Razib Khan @ 11:33 pm

In light of the previous post I was curious about the literature on inbreeding depression of IQ. A literature search led me to conclude two things:

- This is not a sexy field. A lot of the results are old.

- The range in depression for first cousin marriages seems to be on the order of 2.5 to 10 IQ points. In other words ~0.15 to ~0.65 standard deviation units of decline in intelligence.

The most extreme case was this paper from 1993, Inbreeding depression and intelligence quotient among north Indian children. The authors compared the children of first cousin marriages, and non-bred in individuals, from a sample of Muslims in Uttar Pradesh of comparable socioeconomic status (though the authors note that inbreeding has a positive correlation with socioeconomic status in this community). A table with results speaks for itself:


May 29, 2012

Reason: the God that fails, but we keep socially promoting….

Filed under: Anthroplogy,Cognitive Science,Psychology — Razib Khan @ 12:03 pm

One point which I’ve made on this weblog several times is that on a whole range of issues and behaviors people simply follow the consensus of their self-identified group. This group conformity probably has deep evolutionary origins. It is often much cognitively “cheaper” to simply utilize a heuristic “do what my peers do” than reason from first principles. The “wisdom of the crowds” and “irrational herds” both arise from this dynamic, positive and negative manifestations. The interesting point is that from a proximate (game-theoretic rational actor) and ultimate (evolutionary fitness) perspective ditching reason is often quite reasonable (in fact, it may be the only feasible option if you want to “understand,” for example, celestial mechanics).


If you’re faced with a complex environment or set of issues “re-inventing the wheel” is often both laborious and impossible. Laborious because our individual general intelligence is simply not that sharp. Impossible because most of us are too stupid to do something like invent calculus. Many people can learn the rules for obtaining derivatives and integrals, but far fewer can come up with the fundamental theorem of calculus. Similarly, in the 18th century engineers who utilized Newtonian mechanics for practical purposes were not capable ...

Reason: the God that fails, but we keep socially promoting….

Filed under: Anthroplogy,Cognitive Science,Psychology — Razib Khan @ 12:03 pm

One point which I’ve made on this weblog several times is that on a whole range of issues and behaviors people simply follow the consensus of their self-identified group. This group conformity probably has deep evolutionary origins. It is often much cognitively “cheaper” to simply utilize a heuristic “do what my peers do” than reason from first principles. The “wisdom of the crowds” and “irrational herds” both arise from this dynamic, positive and negative manifestations. The interesting point is that from a proximate (game-theoretic rational actor) and ultimate (evolutionary fitness) perspective ditching reason is often quite reasonable (in fact, it may be the only feasible option if you want to “understand,” for example, celestial mechanics).


If you’re faced with a complex environment or set of issues “re-inventing the wheel” is often both laborious and impossible. Laborious because our individual general intelligence is simply not that sharp. Impossible because most of us are too stupid to do something like invent calculus. Many people can learn the rules for obtaining derivatives and integrals, but far fewer can come up with the fundamental theorem of calculus. Similarly, in the 18th century engineers who utilized Newtonian mechanics for practical purposes were not capable ...

May 16, 2012

Bias in psychology

Filed under: Psychology,science — Razib Khan @ 9:46 pm

Ed Yong has a piece in Nature on the problems of confirmation bias and replication in psychology. Yong notes that “It has become common practice, for example, to tweak experimental designs in ways that practically guarantee positive results.” The way this has been explained to me is that you perform an experiment, get a p-value of > 0.05 (significance). You know that your hunch is warranted, so just modulate the experiment, and hope that the p-value comes in at < 0.05, and you have publishable results! Obviously this is not just a problem in psychology; John Ioannidis has famously focused on medicine. But here’s a chart which shows that positive results are particular prevalent in psychology:

There are many angles to this story, but one which Ed did not touch upon is the political homogeneity of of psychology as a discipline. The vast majority of psychologists are political liberals. This issue of false positive results being ubiquitous is pretty well known within psychology, so I’m sure that that’s one reason Jonathan Haidt has emphasized the ideological blinders of scholars so much. Let’s assume that the ...

January 5, 2012

The “sex difference factor”?

Filed under: Psychology,Sex Differences — Razib Khan @ 3:10 pm

There’s a new paper in PLoS ONE, The Distance Between Mars and Venus: Measuring Global Sex Differences in Personality*, which suggests that by measuring variation of single observed personality traits researchers are missing larger underlying patterns of difference. The Distance Between Mars and Venus: Measuring Global Sex Differences in Personality:

In conclusion, we believe we made it clear that the true extent of sex differences in human personality has been consistently underestimated. While our current estimate represents a substantial improvement on the existing literature, we urge researchers to replicate this type of analysis with other datasets and different personality measures. An especially critical task will be to compare self-reported personality with observer ratings and other, more objective evaluation methods. Of course, the methodological guidelines presented in this paper can and should be applied to domains of individual differences other than personality, including vocational interests, cognitive abilities, creativity, and so forth. Moreover, the pattern of global sex differences in these domains may help elucidate the meaning and generality of the broad dimension of individual differences known as “masculinity-femininity”…In this way, it will be possible to build a solid foundation for the scientific study of psychological sex differences and their biological and cultural origins.


I’m curious about the reaction of people in psychology to this result. The reason is that I am generally confused or skeptical about measurements of personality difference. I’m not confused or skeptical of differences in personality between individuals or groups. I agree that these exist. I just don’t have a good sense of the informativeness of the measures of difference. People may criticize psychometrics intelligence testing all they want, but at least their methods are relatively clear.

From what I can gather the authors discovered that the differences between sexes on personality were much clearer once you looked for the correlation across numerous single measured traits. This strikes me as similar to what you see in population genetics when you move from variation in one gene across populations to many. While a single gene is not very informative in terms of population differences (e.g., the standard assertion that ~15 percent of variation is between races), synthesizing the variation of many genes allows one to easily distinguish populations, because there is such strong discordance in the correlation of differences. An analogy with traits makes understanding this easy. If you were told that population X tended toward black hair, that would not be very informative. Nor if you were told that population X tended toward straight hair. And what if you were told that population X tended toward light skin? All these traits are common across many different populations. But if you told that population X tended toward straight black hair and light skin, the set of populations which intersect at those three traits together in this direction is far smaller than evaluating on a trait-by-trait basis.

But in regards to the evolution of sex differences there is something that I feel that I can say here. Humans seem to lay between other ape lineages in terms of physical dimorphism. For example, in size the difference between males and females is not as extreme as gorillas, but not as equitable as among gibbons. These differences are traditionally correlated with social structure. Groillas are highly polygynous, and there is a great deal of male-male competition, therefore driving sexual selection. In contrast, gibbons tend toward monogamy (at least in the ideal, as with “monogamous birds” the reality seems to differ from the ideal).

But there is also an evolutionary genetic aspect to sexual dimorphism we must consider: in Genetics and Analysis of Quantitative Traits the authors note that evolution of sex specific traits is not going to occur fast. The reason is simple: aside from the peculiarities on the sex chromosomes males and females are genetically the same. This implies that sex differences on the genetic level may emerge via modulation of gene expression across networks of genes tuned by some “master controllers” associated with differential sex development. All of this added complexity takes time to evolve, with the rough result that sexual differences in trait value take about an order of magnitude longer than other traits to come to the fore. The intuition here is simple: if there is selection for large males, there will be selection for large daughters indirectly. Modifiers which dampen this effect need to emerge, so that sex-specific selection doesn’t have the side effect of dragging the other sex along in terms of trait value (this is a concern when you have traits, such as high testosterone, which might increase fitness in males, but reduce it their daughters). Therefore, if there are sex differences in behavioral tendencies which are biologically rooted (I believe there), they will tend to be universal across human societies and have a very deep evolutionary history.

So that would be the strategy to understand differences in personality across the sexes. Go beyond W.E.I.R.D. populations, as they did in this study. And look for traits where males and females seem to exhibit consistent differences across these range of social environments. I suspect environment does effect the magnitude of differences, but I would be willing to bet money that some differences are going to persist (e.g., inter-personal violence is an area where males will differ due to size and personality).

* I’m really sick of the use of the Mars vs. Venus dichotomy in the scholarship.

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?

December 27, 2011

Richard Feynman’s intelligence

Filed under: Psychology,Richard Feynman — Razib Khan @ 1:32 am

Interesting interview of Steve Hsu. I’ll reproduce the part about Feynman:

3. Is it true Feynman’s IQ score was only 125?

Feynman was universally regarded as one of the fastest thinking and most creative theorists in his generation. Yet it has been reported-including by Feynman himself-that he only obtained a score of 125 on a school IQ test. I suspect that this test emphasized verbal, as opposed to mathematical, ability. Feynman received the highest score in the country by a large margin on the notoriously difficult Putnam mathematics competition exam, although he joined the MIT team on short notice and did not prepare for the test. He also reportedly had the highest scores on record on the math/physics graduate admission exams at Princeton. It seems quite possible to me that Feynman’s cognitive abilities might have been a bit lopsided-his vocabulary and verbal ability were well above average, but perhaps not as great as his mathematical abilities. I recall looking at excerpts from a notebook Feynman kept while an undergraduate. While the notes covered very advanced topics for an undergraduate-including general relativity and the Dirac equation-it also contained a number of misspellings and grammatical errors. I doubt Feynman cared very much about such things.


One thing I have always wondered about is the fact that Richard Feynman had substantive accomplishments which marked him as definitively brilliant by the time he was talking about his 125 I.Q. score (which is smart, but not exceedingly smart). Intelligence scores are supposed to be predictors of accomplishments, but Feynman already had those accomplishments. Bright people take many psychometric tests, so there will be a range of score about a mean. My personal experience is that there’s a bias in reporting the highest scores. But it may be that Feynman gloried in reporting his lowest scores because that made his accomplishments even more impressive. Unlike most he had nothing to prove to anyone.

December 26, 2011

The aliens among us

Filed under: Asperger syndrome,Psychology — Razib Khan @ 3:48 pm

Amy Harmon has a very long piece in The New York Times, Navigating Love and Autism. It’s about a couple who both have been diagnosed with Asperger syndrome. Like cancer I suspect that this term brackets a lot of different issues into one catchall label, not to mention the acknowledgment that it’s a spectrum. When I spent time with the Bay Area Less Wrong community I would observe the range in tendencies and neurological diversity of people who clearly would be classified as “high functioning autistic” (to be clear, these were individuals strongly selected for high general intelligence, with a minimum threshold of around two standard deviations above the norm). The lack of comprehension of religiosity and bias toward libertarianism were two salient characteristics of this sect (though people who have met me don’t classify me as having Asperger syndrome, I have these two cognitive biases myself)

 

In any case, the bigger issue which Amy Harmon’s piece brought out to me is that people with high-functioning autism develope their own micro-norms, meaning that they are often not very compatible with each other despite their deviation from “neurotypicals.” There’s no guarantee that you’ll deviate away from the norm in the same dimension when the norm is highly multidimensional!

People with Asperger are often non-conformists. This is not a bad thing necessarily, at least for society as a whole. But as explained in Not by Genes Alone: How Culture Transformed Human Evolution a very strong tendency toward within-group conformity is a major hallmark in human behavior. It’s probably biological encoded. So, for example, speaking with your parents’ accents, as opposed to that of your hypothetical peer group, is a trait of many people with high functioning autism (or, a tendency toward hyper-formalism of speech). This is a tell for lack of group conformity. The problems autistic people have with conventional “manners,” and not just basic universal human niceties, is an outgrowth of this tendency I suspect. Manners can differ greatly across societies, and require cultural conditioning. But the human tendency to want some set of regular norms does apply to those with Asperger. The diversity among this set is what results in the difficulties of negotiating conflicts (and may explain a bit why libertarians and the hyper-atheistic tend to fracture along what seem trivial deviations from the outside!). You can imagine that in some ways people with Asperger syndrome explore the full parameter space of cultural possibilities, unencumbered by the positive feedback loops of group conformity which is the human norm.

December 11, 2011

Most Reported Genetic Associations with General Intelligence Are Probably False Positives

Filed under: Behavior Genetics,Genomics,Human Genomics,I.Q.,Psychology — Razib Khan @ 4:49 pm

The title says it all, and I yanked it from a paper that is now online (and free). It’s of interest because of its relevance to the future genetic understanding of complex cognitive and behavioral traits. Here’s the abstract:

General intelligence (g) and virtually all other behavioral traits are heritable. Associations between g and specific single-nucleotide polymorphisms (SNPs) in several candidate genes involved in brain function have been reported. We sought to replicate published associations between 12 specific genetic variants and g using three independent, well-characterized, longitudinal datasets of 5571, 1759, and 2441 individuals. Of 32 independent tests across all three datasets, only one was nominally significant at the p ~ .05 level. By contrast, power analyses showed that we should have expected 10–15 significant associations, given reasonable assumptions for genotype effect sizes. As positive controls, we confirmed accepted genetic associations for Alzheimer disease and body mass index, and we used SNP-based relatedness calculations to replicate estimates that about half of the variance in g is accounted for by common genetic variation among individuals. We conclude that different approaches than candidate genes are needed in the molecular genetics of psychology and social science.


My hunch is that these results will be unsatisfying to many people. The authors confirm and reassert the heritability of general intelligence, both by reiterating classical results, and utilizing novel genomic techniques. But, they also suggest that the candidate gene literature is nearly worthless because of the lack of power of most of the earlier studies. The latter is probably due to the genetic architecture of the trait. Intelligence may be determined by numerous genes of very small effect (e.g., 0.01% of the variance effected by one particular SNP), or, “rare, perhaps structural, genetic variants with modest to large effect sizes.” The former case is pretty obvious, but what about the latter? I’m mildly skeptical of this because I’m curious why modest-to-large effect variants didn’t show up in family-based studies (presumably within the family the same variants would localize to sections of the genetic map)? But I’m not fluent enough in the literature to know if there was a lot of work in this area with families previously.

Related: Here’s the first author’s article in Commentary from the late 1990s, IQ Since “The Bell Curve”.

November 28, 2011

The New York Times on violence and Pinker

The New York Times has a short piece on Steven Pinker up. Nothing too new to long time followers of the man and his work. I would like to point readers to the fact that Steven Pinker has a F.A.Q. up for The Better Angels of Our Nature: Why Violence Has Declined. He links to my post, Relative angels and absolute demons, as supporting his dismissal of Elizabeth Kolbert’s review in The New Yorker. I have to admit that I find much, though not all, of the coverage of science in The New Yorker to exhibit some of the more annoying stereotypical caricatures of humanists when confronting the specter of natural philosophy.

I should also mention I started reading The Better Angels of Our Nature over Thanksgiving. I’m only ~20% through it, and probably won’t finish until Christmas season gets into high gear, but so far it’s a huge mess. In both a good way, and a bad way. The good way is that it’s incredibly rich in its bibliography, with fascinating facts strewn about the path of the narrative. The bad way is that so far it lacks the tightness of  The Blank Slate or The Language Instinct in terms of argument. This may change. Finally, I think I should mention that Pinker has already addressed some of the criticisms of his methodologies brought up in the comments sections of my posts. Those who have specific critiques probably should read the book, because he seems to try sincerely to address those. Or at least they should address those critiques to people who have bothered to read the book.

November 10, 2011

The problem of false positives

Filed under: Psychology — Razib Khan @ 7:01 pm

False-positive psychology: undisclosed flexibility in data collection and analysis allows presenting anything as significant:

In this article, we accomplish two things. First, we show that despite empirical psychologists’ nominal endorsement of a low rate of false-positive findings (≤ .05), flexibility in data collection, analysis, and reporting dramatically increases actual false-positive rates. In many cases, a researcher is more likely to falsely find evidence that an effect exists than to correctly find evidence that it does not. We present computer simulations and a pair of actual experiments that demonstrate how unacceptably easy it is to accumulate (and report) statistically significant evidence for a false hypothesis. Second, we suggest a simple, low-cost, and straightforwardly effective disclosure-based solution to this problem. The solution involves six concrete requirements for authors and four guidelines for reviewers, all of which impose a minimal burden on the publication process.

Since the paper is behind a paywall, I’ve cut & pasted the solutions belows:

We propose the following six requirements for authors.

  1. Authors must decide the rule for terminating data collection before data collection begins and report this rule in the article. Following this requirement may mean reporting the outcome of power calculations or disclosing arbitrary rules, such as “we decided to collect 100 observations” or “we decided to collect as many observations as we could before the end of the semester.” The rule itself is secondary, but it must be determined ex ante and be reported.

  2. Authors must collect at least 20 observations per cell or else provide a compelling cost-of-data-collection justification. This requirement offers extra protection for the first requirement. Samples smaller than 20 per cell are simply not powerful enough to detect most effects, and so there is usually no good reason to decide in advance to collect such a small number of observations. Smaller samples, it follows, are much more likely to reflect interim data analysis and a flexible termination rule. In addition, as Figure 1shows, larger minimum sample sizes can lessen the impact of violating Requirement 1.

  3. Authors must list all variables collected in a study. This requirement prevents researchers from reporting only a convenient subset of the many measures that were collected, allowing readers and reviewers to easily identify possible researcher degrees of freedom. Because authors are required to just list those variables rather than describe them in detail, this requirement increases the length of an article by only a few words per otherwise shrouded variable. We encourage authors to begin the list with “only,” to assure readers that the list is exhaustive (e.g., “participants reported only their age and gender”).

  4. Authors must report all experimental conditions, including failed manipulations. This requirement prevents authors from selectively choosing only to report the condition comparisons that yield results that are consistent with their hypothesis. As with the previous requirement, we encourage authors to include the word “only” (e.g., “participants were randomly assigned to one of only three conditions”).

  5. If observations are eliminated, authors must also report what the statistical results are if those observations are included. This requirement makes transparent the extent to which a finding is reliant on the exclusion of observations, puts appropriate pressure on authors to justify the elimination of data, and encourages reviewers to explicitly consider whether such exclusions are warranted. Correctly interpreting a finding may require some data exclusions; this requirement is merely designed to draw attention to those results that hinge on ex post decisions about which data to exclude.

  6. If an analysis includes a covariate, authors must report the statistical results of the analysis without the covariate. Reporting covariate-free results makes transparent the extent to which a finding is reliant on the presence of a covariate, puts appropriate pressure on authors to justify the use of the covariate, and encourages reviewers to consider whether including it is warranted. Some findings may be persuasive even if covariates are required for their detection, but one should place greater scrutiny on results that do hinge on covariates despite random assignment.

Guidelines for reviewers

We propose the following four guidelines for reviewers.

  1. Reviewers should ensure that authors follow the requirements. Review teams are the gatekeepers of the scientific community, and they should encourage authors not only to rule out alternative explanations, but also to more convincingly demonstrate that their findings are not due to chance alone. This means prioritizing transparency over tidiness; if a wonderful study is partially marred by a peculiar exclusion or an inconsistent condition, those imperfections should be retained. If reviewers require authors to follow these requirements, they will.

  2. Reviewers should be more tolerant of imperfections in results. One reason researchers exploit researcher degrees of freedom is the unreasonable expectation we often impose as reviewers for every data pattern to be (significantly) as predicted. Underpowered studies with perfect results are the ones that should invite extra scrutiny.

  3. Reviewers should require authors to demonstrate that their results do not hinge on arbitrary analytic decisions. Even if authors follow all of our guidelines, they will necessarily still face arbitrary decisions. For example, should they subtract the baseline measure of the dependent variable from the final result or should they use the baseline measure as a covariate? When there is no obviously correct way to answer questions like this, the reviewer should ask for alternatives. For example, reviewer reports might include questions such as, “Do the results also hold if the baseline measure is instead used as a covariate?” Similarly, reviewers should ensure that arbitrary decisions are used consistently across studies (e.g., “Do the results hold for Study 3 if gender is entered as a covariate, as was done in Study 2?”).5 If a result holds only for one arbitrary specification, then everyone involved has learned a great deal about the robustness (or lack thereof) of the effect.

  4. If justifications of data collection or analysis are not compelling, reviewers should require the authors to conduct an exact replication. If a reviewer is not persuaded by the justifications for a given researcher degree of freedom or the results from a robustness check, the reviewer should ask the author to conduct an exact replication of the study and its analysis. We realize that this is a costly solution, and it should be used selectively; however, “never” is too selective.

To preempt angry and offended psychology professors: this problem is not limited to their discipline. It is probably a bigger problem in medicine because it costs us a lot of money and likely kills people.

Older Posts »

Powered by WordPress