It's my first time here, so sorry if I'm too direct.
I'm dealing with a 0-1 variable that I want to model (logistic regression) but sample sizes are very unbalanced (2489 data, only 58 are positive) so models can converge. I was thinking about the possibility of randomly sampling 58 negative cases, so I can run a balanced logistic regression. I did it, n-times, and I wonder how I can "average" my results. One option, with univariate models, is simply show the CI for the estimates obtained from the n-models obtained, but my question is how to deal with models including more than one predictor. There is an option in R to do that? or manually, how can I do it (I can do it as with univariate models, but I'm not sure if this is fine)? How to decide which explanatory variables are actually contributing to explain variability in the dependent variable? any advice is welcome.