I'm currently wrapping up my dissertation and the final chapter involves using twins and siblings to disentangle genetic and environmental influences on the associations between sleep health and psychiatric disorders.
One of the analyses is a form of a co-twin control model. Twins can be used as "natural experiments" in behavioral sciences because they either share 50% (fraternal or dizygotic) or 100% (identical or monozygotic) of their genetics on average, and we assume they share 100% of their common / familial / rearing environment. I've been working on writing up the results for this project and after watching the super bowl was thinking about how there are a decent amount of siblings in the NFL -- as well as in other pro sports.
Sports media outlets love to say someone has the "clutch" gene or the "basketball" gene etc, and while that definitely isn't true (there isn't one gene to explain pretty much any complex behavioral trait, but instead hundreds to thousands of genetic variants that have tiny effect sizes), you also have to consider the environment.
Did Serena and Venus Williams inherit some sort of genetic combination that predisposed them to be great at tennis, or was their Dad heavily involved in tennis and did he have them spending hours on hours at tennis courts growing up? Most likely a combination of both! That is the case with most complex human genetics -- both genetics and environment contribute to the variance in behavioral traits.
So today I did a small analysis looking at recent siblings in the NFL. The full code and write up is here: <https://rpubs.com/clairemo/NFL_sibs>, but briefly, I did a within-sibling analysis that looks at genetic and environmental effects on the relationship between draft selection order and pro football focus (PFF) grade. Here are the variables I pulled out from the PFF website and how they correlate:
"PFF" is their PFF ranking number at the end of their first season played, "speed" is their 40 yard dash time at draft year, "draft" is the number they were selected in their draft class, and "PFF_yr_3" is their PFF ranking 3 years after being drafted.
In a mixed model where I let the average draft value for each family and each player's deviation from that average draft number predict PFF grade in year 3, the deviation variable (or within-sibling effect) was significant. What this means is the higher (earlier) draft pick you are, the more likely you are to have a higher (better) PFF grade in year 3. Specifically, as your draft pick number decreases 1, we expect a 0.08 increase in PFF grade at year 3.
Further, this gives us some confidence that an exposure pathway exists between draft number and PFF grade in year 3. We can be somewhat confident in that statement because the siblings act as near perfect controls for each other. When the deviation independent variable significantly predicts the dependent variable, it is implying that what is predicting the dependent variable (someone's PFF grade in year 3) is how different they are from their average family draft number. In other words, because we are looking at this deviation variable, we are controlling for familial factors that make siblings more similar, such as nutrition, parenting, etc. And because the difference in draft order still significantly predicts PFF grade at year 3, there is the potential for an exposure pathway between draft order and PFF grade at year 3 (controlling for genetics and familial environment).
(If the average family draft order variable significantly predicted PFF grade in year 3 we would interpret that result as familial factors influencing draft order predicting PFF grade in year 3.)
You can see that exhibited in the plot below by siblings like the Kelce's, McCourty's or Bennetts -- across the x axis, those siblings were selected at quite different points in the draft. But despite that, they still have pretty high PFF grades in year 3.
What is predicting Travis Kelce's PFF grade, for example, is how different his draft selection number was from the average of his and his brother Jason's draft selection number. We can't fully rule out genetics here because Travis and Jason only share, on average, 50% of their genetic influences, so they are not perfectly matched controls. But we are attempting to rule out other unmeasured factors that might contribute to football success like playing competitively, being around the game, money spent on private coaches, nutrition etc. The green square represents those who were high (earlier; top 50) draft picks and also maintained a high PFF (>70) in year 3.
Of course it is also highly important to note the tiny sample size here -- but overall this was just a fun analysis to distract me from endless dissertation work!
Comments