Correlation And Weighted Percentages
1
2
Entering edit mode
9.4 years ago
dmw610 ▴ 20

This is more of a general statistics question, but I've gotten awesome help here before so I figured I'd try it out again.

I noticed that in my department, older professors seem to have more male-dominated labs than younger professors do. I collected the ages and gender makeup of 100 labs, and there is indeed a small correlation between the age of the professor and the gender of his or her employees (Pearson's r = .27). The problem is that the professors all have different numbers of workers. One professor might have 1 male and 0 female employees, while another professor has 14 male and 2 female employees. But calling the first lab "100% male" and the second "87.5% male" obscures the fact that the gender bias is much worse in the second lab. I want to do correlation analysis that takes into account the different sizes of the labs. Does anyone have any thoughts on how to do this? Thanks!

correlation • 1.8k views
0
Entering edit mode

Two important statistical points that I've learned: One, is your sample a random sample? Statistics are only valid given a random sample unless you have access to the whole population. Two, was the correlation (Pearson r=0.27) significant? What was the p-value you obtained? I'm eager to also expand on these learnings. Any pro to pitch in?

1
Entering edit mode
9.4 years ago

It sounds like what you really want is a logistic regression. This is relatively easy in R and here's a nice tutorial from UCLA.