Thursday, April 21, 2022

Years of Educational Attainment

This is becoming a more common thing to measure in genetics over the last decade or so. There is a huge advantage in being able to just plain get ahold of the data, as people fill it out on profiles of all sorts.  From various ancestry sites you can get millions of answers.  On Unsupervised Learning, Khan interviews UMinn psychology prof James Lee about his recent paper in Nature. It has a GWAS of 3 million individuals. 

This seems intuitively weak at first, because a lot of things go into educational attainment that aren't genetic, so what good is it? What country you were born in or moved to, what your family's expectations are, what your chosen field requires, your level of determination (even if that is also genetic) all factor in to that final number. Even the final number can be fuzzy.  Someone who enters "13" might have been dragged across the finish line for high school and stuffed into a college for a couple of semesters picking up only a few credits.  Or they might have left college to go into the service or a particular job where they received ongoing training for many years.  The life-outcomes of the individual might be very different with that number. However, in large studies like this the researchers are mostly just worried about the range. For both 10 and 11 years, or 13 and 14 years, the +/- might be the same two years, so it washes out. Only at the graduation points of 12 and 16 is there likely to be a bending of the graph as lots of individuals try to get those last few classes in to get to the finish line.

The genetic correlation for height is 91%, for IQ 83%, and for years of educational attainment about 55%. These seem about right.  You can't make yourself taller, but your height can get screwed up, such as by backbone curvature, as my son from Romania's is, making him about 2" shorter. Similarly, even more things can work to depress your IQ, though you can't stretch it above a certain point. And for educational attainment, there are many reasons that can get bounced around, as noted above. Still, we would expect that there was some genetic association.  Of particular interest is that Lee notes that the genetic differences cluster around neurological development, which also makes intuitive sense. 

The long interview included some very clear descriptions about why heterozygotic and homozygotic - that is, whether you have a plus and a minus gene at any given point or whether you have two pluses or two minuses - have different effects than expected.  We are familiar with the example of the Sickle-Cell Anemia genes, where having no copies of the immunity is bad, a copy from only one parent is good, and having a copy from both parents is very bad. We were taught years ago in biology that lots of genes behave this way.  In fact, very few do. They are interesting because these rarer genes are often the ones with strongest effects, mostly negative but also some positive.


David Foster said...

"The genetic correlation for height is 91%, for IQ 83%, and for years of educational attainment about 55%. These seem about right."

Maybe this is true if you look at a snapshot in time, but it is not true over long time spans. Nutrition has a huge influence on height.

Assistant Village Idiot said...

My recollection is that it is about early childhood and very dramatic differences in nutrition, such as prolonged severe undernourishment in wartime. IIRC, it was about the Dutch being short if very young after WWI but becoming the tallest people in the world by the 1980s. I'll see if I can find more.