Is there a good way to test whether explanatory power among variables varies across populations?

[Jim Johnson]

Let's imagine I want to assess whether the explanatory power of daily calories, minutes of exercise, and hours of sleep has different explanatory power over BMI for men compared to women. Is there a good way to evaluate this beyond just doing regressions for each sample and comparing R^2?

I could run two models and compare their R-squared values, but I'm wondering if there is a better way.

[Carlo Lazzaro]

Jim:
an interaction would probably do the trick:

[Jim Johnson]

Interactions would tell me whether the effect size differs, but can they tell me whether explanatory power differs?

[Carlo Lazzaro]

Jim:
see Example 14: Calculating effect sizes for a linear regression model, -regress postestimation- entry, Stata .pdf manual.