hello,
I have a short panel (t=3 and N=2000), the main model I estimated was an ordered probit as the dependent variable was an ordered categorical variable. But a regressor I have a way of thinking is endogenous. So I estimate the model with the cmp (Roodman) approach also because the instrumental variable I have is a binary variable.
To consider fixed effects I read that it is necessary to put a dummy for each cross-sectional unit in the model, unfortunately STATA processes hours and hours but I do not get any results.
I tried to calculate the fixed effects manually, but I don't get any result.
THE PROBLEM I think is the fact that the 3 waves were carried out on the same sample after one month and therefore many variables do not vary.
the alternative would be to estimate a cmp with random effects, but how could I justify the correlation between the unobserved effects and the regressors?
Thank you
I have a short panel (t=3 and N=2000), the main model I estimated was an ordered probit as the dependent variable was an ordered categorical variable. But a regressor I have a way of thinking is endogenous. So I estimate the model with the cmp (Roodman) approach also because the instrumental variable I have is a binary variable.
To consider fixed effects I read that it is necessary to put a dummy for each cross-sectional unit in the model, unfortunately STATA processes hours and hours but I do not get any results.
I tried to calculate the fixed effects manually, but I don't get any result.
THE PROBLEM I think is the fact that the 3 waves were carried out on the same sample after one month and therefore many variables do not vary.
the alternative would be to estimate a cmp with random effects, but how could I justify the correlation between the unobserved effects and the regressors?
Thank you
Comment