Logistic Regression: Evaluate model & Check for Multicollinearity


I have conducted a logistic regression with a binary categorical outcome and a binary categorical moderator + binary categorical independent variable (or more specific: two independent variables and in the logistic regression entered the interaction factor).
How do I now check how good the model fits the data (Nagelkerke?), if I need mean-centering (all my variables are coded into 0 and 1 as values)? I have also read others checking for standardized residuals and Cooks Distance but I would like to find out if it is needed in my case. Also, do I need to check for multicollinearity in the logistic regression? Or would that only be in a linear regression?


Less is more. Stay pure. Stay poor.
You do not need to center your categorical variables. Pseudo-R^2 metric are frown on. People like to look at calibration, Hosmer-Lemeshow values (fitted vs. expected) and c-statistics.


Less is more. Stay pure. Stay poor.
I would do a internet search and include whichever software you are using as a key term. C-stat is also called accuracy or concordance statistic.