[Limdep Nlogit List] Heteroscedastic Conditional logit for Panel data

[Dan Rigby]

The Error Components Model seems to introduce the possibility of allowing the variance/scale parameter to vary across two samples (i.e. SP and RP) but does so for each alternative within the choice set separately.  

However, in an unlabeled choice experiment one may want the error variances to be constant across alternatives but to vary across samples/people 

(i.e. there is no reason to believe the error variance will be different within the choice question).

 

I want to do this while accounting for the panel (repeated choice) nature of the choice data. 

Is this possible within NLOGIT?

 

Thanks,

Dan Rigby