Dear all,
I am running a rank ordered probit model (cmroprobit in Stata 16), in which I have a sample of around 300 individuals that rank the preference for 4 different attributes. There are not alternative-specific variables but case-specific variable (related to respondents’ characteristic), mostly categorical variables. I have two questions:
  1. Does the rule 10 obs. per variable holds also for the rank ordered probit?As the sample is quite small (300), and each variable have 3 parameters (four alternatives one of which is the base alternative), I should use no more than 10 variables.
  2. Each individual repeat the rank-ordered choice three times, under three different conditions. I was so wondering whether it is possible in Stata to model the rank-ordered probit in a panel data fashion. From the “choice models” manual in Stata 16, it seems that only the mixed-logit model can be modelled in panel data.Do you confirm? and if so, do you have any suggestions?
Thank you