This is a weird thing to post in HC, but I thought I'd try to tap into the knowledge base that often gathers here and see if anyone has used Negative Binomial Regression in hypothesis testing (differences between groups).
I conducted an experiment and have two groups exposed to different treatment conditions and then measured performance on a problem-solving task. My data is counts (number errors) and hence is zero-inflated and after lots of consideration I have identified that it best fits a negative binomial distribution. I had originally intended to use ANCOVA to address my primary RQ (effect of the IV levels of treatment condition on the number of errors on the task), and also have covariates of interest (pre and post measure of emotion states). Ancova normality assumption is violated (Shapiro-Wilks was significant), and so I wonder if is even appropriate to proceed with ANCOVA given this variable has non-normal, zero inflated distribution.
Could negative binomial regression provide an equivalent analysis for hypothesis testing in such a case? ANCOVA and Regression are all in the GLM family, but am unsure if regression appropriate for experimental hypothesis testing.
Hoping someone knows what I am talking about here and can offer some advice or feedback.
