I'm rusty on my stats. I am wondering what the prerequisites are for including covariates in my regression. I'm not looking for a prediction model but just the effect of the "treat". The sample was randomized and the demographics appear pretty similar with a couple of exceptions.
Do I include all or only those that are the same in both treatment groups? Do I need to first do a statistical test to compare the two groups?
Few of the demographics appear to make a significant difference in the regression, but if I exclude them this hurts the significance level of the treat coefficient.