Conditional logit (with a different number of alternatives for each case)?

#1
Dear all,

I have a question about conditional logit analysis, in particular analysis based on a McFadden-type choice model. I am using a model of firm relocation destination county choice using a 19th century panel dataset, with the main explanatory variables being county-specific variables. Since this is the 19th century, the choice of county alternative destinations is different for firms relocating in different years since some counties weren't in existence in some years and some were merged with others in other years. Is it inappropriate for cases (firms) to have a different number of alternatives? In other words, can I exclude some destination alternatives (counties) for some cases/firms because those counties were not in existence in particular years?

Also, I am using cmclogit in State for this analysis. Would cmmixlogit be a more appropriate for these kind of data?

Thank you so much, in advance!

Chirag