I am working on an assignment involving a logistic regression model, where I need to plot the pearson standardized residuals against one of the predictors. Here's the basic setup:
model <- glm(outcome ~ predictor1 + predictor2, family=binomial(logit)) res <- residuals(model, "pearson")
When looking at the residuals' distribution, I see something totally different than my colleagues who use Stata (using predict and rstandard). Their residuals are more or less normal, whereas in mine there is a gap in the values (not a singe residual is between -0.05 and 1.15). That does make sense in the context of logistic regression, especially that the maximum predicted probability is not so high (38%).
I'd like to understand what's happening here... What is Stata doing that R isn't, with those residuals?