Multinomila logit with different sets of alternative specific variables

Hi.

I have a question regarding multinomial logit models in STATA. I'm trying to estimate a trip mode choice model in stata. The alternatives have different specific variables. For Instant the transit mode have repressors such as waiting time or interchange time while these are not applicable for walking mode and bicycle mode or car mode.

I have looked at modules "asclogit" and "nlogit" and it seems to me that the alternative specific variables need to be same variable for all alternatives (in my case modes) while my model does not allow that.



I also have tried to set the value of the variables to zero when they were not applicable for the specific modes which results in error message: “variable xxx is not alternative specific: it has not within-case variability"
Am I doing something wrong here?
I also have same type of question but in case of mlogit.
If I model a multinomial logit and for some alternatives, some of the repressors become insignificant, I would like to exclude them from the final model. Is there a way of doing this or am I making a mistake here?