I have data from the UK National Travel Survey and I am estimating McFadden's
choice model. I have three different modes of transport, choice specific
attributes (time, cost, speed) and individual characteristics (age,
sex, income, etc).
I have estimated the model using clogit given that I have variables that vary
across the choices.
After rearranging the data I have estimated the multinomial logit, mlogit,
without the choice specific attributes and I obtained the same results as
the clogit ommiting them as well.
My first question:
I have done the estimation with the mlogit including the choice specific
attributes and Stata did not crash. My guess is that the results are
meaningless but I wanted to confirm that.
Second:
If I want to estimate a model where the coefficients of time, cost and speed
(the choice specific variables) vary across choices (an unrestricted version
of the clogit model), is it ok if I use clogit with the data in the long
format and interact the choice specific variables by the choices?
Third (a wish):
It would be nice if clogit would compute clustered standard errors as well.
I have done bootstrap to correct my standard errors and it takes ages!
Thanks a lot for your help,
Teresa Delgado
Department of Economics
University College London
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/