1

My dependent variable is a binary outcome(Y); therefore, I am using logistic regression. I am interested in what proportion of variance of Y is explained by the variable X1. In other words, I am interested in the unique variance of Y explained by X1 after controlling for X2 and X3.

This is the concept of partial R-squared in OLS. But I am not sure if there is an equivalent way or statistic I can use for logistic regression. Alternatively, should I use a linear probability model to obtain partial R-squared?

Hyojin
  • 11
  • 1
  • 1
    You could probably use the change in pseudo-R2 corresponding to removal of each predictor. Not sure if there is a straightforward formula for that, and it would depend on which pseudo-R2 you would use. – Noah Oct 08 '21 at 20:42
  • See https://stats.stackexchange.com/questions/3559. A [site search](https://stats.stackexchange.com/search?tab=votes&q=logistic%20regression%20R%20squared) turns up plenty of other relevant threads. – whuber Oct 08 '21 at 21:02

0 Answers0