2
$\begingroup$

I have a panel data (all listed companies in the US with quarterly data) in which the dependent variable is 1 if the company was recommended on a investment recommendations website, and 0 otherwise. The independent variables are several stock characteristics (e.g. ROE, P/E, sales growth, market capitalisation etc.).

I want to run a logit regression (using pglm package in R) but I'm not sure if I should resample or not. The dependent variable is 1 only in 2.3% of the cases. If I run the regression without resampling, I get significant results but also the intercept is very high (-3.95 at a t value of -134.5). How do you think I should approach this?

$\endgroup$
0
$\begingroup$

You SHOULD NOT resample to balanced the 0 and 1 since it will bias the result.

When you run the regression, you care more about when you change some variables, whether they will change the outcome, which is the meaning of the coefficients, but not the level of the outcomes (whether the outcomes are all 0, or 1).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.