A couple years back you posted this (see below) message -- I am extremely interested in the your solution to this problem. I am estimating a dynamic multinomial logit model and would like to create a transition matrix of the conditional probabilities.
Thank you for the help.
Reka
just wondering whether anyone has coded transition models, also known as
> Markov models, for 3 or more variables. That is, say a system has 3
states.
> Then you want to estimate Pr(y_t+1=1 | y_t=1), Pr(y_t+1=2 | y_t=1), and so
> on. The transition probabilities, given as functions of independent
> variables, need not be identical of course; otherwise it would just a an
> ordered probit or a multinomial logit.
>
> I know that this type of estimation goes by many names, depending on one's
> field, but I haven't been able to find any references to it in any of the
> previous STB's. Any help would be much appreciated.
>
> Thanks,
>
> David Epstein
Reka Sundaram-Stukel
Graduate student
320 Taylor Hall
Department of Agriculture And Applied Economics
Madison, WI-53706
