Regression diagnostics using pglm package / How to perform tests in Random Effects Logit Regression of panel data_

Dear RStudio community,

I hope this query finds you well and I tremendously hope that you can help me out.

I am currently running a logistical regression on panel data for my thesis. My dependent variable is binary i.e. 1 in case a a specific firm action happened in the period or 0 if not. My panel consists of 74 firms. Before excluding missing values (in the independent variable i.e. proxy metrics based on balance sheet information) I come up with 508 observations for 7 periods.

Now coming to my question: When researching how to actual model this with R, the majority of community posts indicated the usage of the „pglm“ package. Please note that I aim to follow a similar approach to an existing study (i.e. random effects logit model). I already tried it out and was able to get statistically significant results.

(1) However, I would like to apply Hausman Test as well, but I am not sure if this available for „pglm“ models. I am not aware of any testing possibilities for the type of model (pooled,, fixed, random-effects). My research did not help my answering this question on my own, if there is any testing possibility using „pglm“. Apparently, ther is no „pglmtest“. Is this just not available or is there any workaround for this?

(2) What I still lack is the possibilities of testing the conditions for logit regression with my dataset. According to my knowledge, I should at least check for multicollinearity (e.g. using VIF), cross-sectional dependence (e.g. Breusch-Pagan-Test and Pesaran Test), linearity assumption and influential values. Can I just use the normal packages for logit regression (lmtest)?

I am very much looking forward to your feedback. Many thanks in advance for your efforts!

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.