3 Comments
User's avatar
Ricco's avatar

I have no view on the underlying substantive debate here. But your logic makes sense to me as an economist. I’ve always been confused about why there isn’t more use of parameter estimates when trying to discern causal impacts of genes vs other factors.

Expand full comment
Ro'i's avatar

Here's another anecdote regarding R-squared. I refereed a paper that included the statement “The R-squared is very high, indicating that our lack of significant results does not result from an insufficient sample to pick up nonzero effects.” The dependent variable was points earned. The experiment included two games, in one of which it was possible to earn more points (think basketball vs. football). The regressions included a dummy for game, which was fully responsible for the high R-squared.

Expand full comment
Emil O. W. Kirkegaard's avatar

This old article makes a similar point regarding effect sizes.

Lubinski, D., & Humphreys, L. G. (1996). Seeing the forest from the trees: When predicting the behavior or status of groups, correlate means. Psychology, Public Policy, and Law, 2(2), 363–376. https://doi.org/10.1037/1076-8971.2.2.363

Expand full comment