I am working with a colleague on her revise and resubmit and the editor and reviewers have asked her to do a negative binomial regression because the standard deviation of her DV exceeds the mean. Her DV is the likelihood of being sexually aggressive (yes or no) and there is a strong floor effect. She wants to test 2 main effects and their interaction on the DV. Her 1st main effect is a manipulation and is made up of 3 groups. Her 2nd main effect is a continuous variable. I have the code for a negative binomial regression with no interactions. How can you test interactions in a negative binomial regression?
As far as the interactions go, you can create two dummy variables from the 3-group variable and multiply then by the continuous variable to create the interaction terms. The covariates in the model are the two dummy variables, one continuous variable and the two interaction variables.
Negative binomial regression is used with count variables. I don't think this is appropriate for a categorical variable.