Interval regression with fixed effects

[paulvonhippel]

I'm looking for a command that implements interval regression with fixed effects and does not necessarily assume normality.

xtintreg only implements random effects. It also assumes normal residuals.

What alternatives are there?

[Stephen Jenkins]

Look at -gintreg- on SSC. I don't quite understand your remarks about 'fixed effects', though. (Please elaborate.) Adding year indicator variables to your regression could presumably prioduce some estimates ... but my conjecture is that they may not have nice statistical properties (unless T is large, not only N).

[paulvonhippel]

Thanks, Stephen. We have two cohorts of children from approximately 35 US states. Our intention is to include fixed effects for each cohort and state. The sample size per state and cohort is approximately 500. It seems to me state and cohort dummies should work well if the functional form of the model is linear and/or the sample size per state and cohort is large, as it is. But this may be uncharted statistical theory....