Hi,

I am estimating a non-linear model by using clogit. My group variable is "industry by year". I also have 2 categorical variables (d1 and d2) that take values of 0, 1, or 2. I want to see the interaction between a continuous variable c1 and d1 and d2. You can see my code below:

clogit y c1##i.d1##i.d2, group(ind_year)

When I have fixed effects in logit models, I look at the number of observations per dummy variable to minimize the bias. However, in a clogit model, I am not sure if I should look at # of observations within each group (industry by year) or total? For example, I have 100 observations where d1=1 and d2=2, and 150 observations where d1=0 and d2=1 but for each industry-year pair, the maximum # of observations is 15. So, I was wondering if I should be concerned about any potential bias? Thanks in advance.

Ulas