In a city, there are 50 neighborhoods. I try fitting a nested logit model for workplace choice and residence choice. The upper level is of workplace. So, there are 50x50=2500 distinct workplace-residence alternatives. Too many alternatives to implement in the regular statistical packages. I notice that a handful of studies tackle this problem via so-called sampling of alternatives, but the way how to program it seems to be hidden to young researchers and not available in the existing packages, including Stata. Is there any researcher in this forum who know to deal with this problem?
↧