Beauty in the Classroom, Part 2

In part one of this case study we concluded that even after adjusting for different sets of plausible confounding variables, there did seem to be an association between an instructor’s appearance and their course evaluations. In part two we will ask whether this effect is different for men and women.

1. For the moment, ignore the confounding variables you identified in part one. Fit a regression model using gender and beauty that allows the effect of the beauty rating to vary between men and women.
2. Based solely on the point estimates from your model above, does appearance have a larger association with course evaluations for men or women?
3. Using the plotModels function from the mosaic package, plot the two regression lines for men versus women. Describe how the gap in expected course evaluations between men and women changes as a function of the beauty score. Hint: for the plotModel function to work correctly by default, you’ll want beauty to be the first variable on the right hand side of the tilde.
4. Using the bootstrap, assess the evidence in the data for different effects of beauty for men and women. Provide a relevant 95% confidence interval. (Hint: First identify the coefficient that governs whether the effects differ, and the value it would have to take for there to be no difference. Is that a plausible value for the interesting coefficient?)
5. Repeat parts 1, 2, and 4, but now using the confounding variables and other predictors you identified in part 1. Do your conclusions change?