Hi Statalisters,
I am using a discrete time multinomial logit with five competing risks, plus non-choice being the sixth outcome. It is pretty clear that I have to deal with the IIA issue somehow. The asmprobit command won't work as I cannot ascribe alternative-specific values in the periods in which no choice is made. So, I was wondering whether multinomial logit with random effects, as it is describe in the gllamm manual, can be of help here? In other words, does it help me relax the IIA assumption and to what extent? Many thanks!
Dimitrije
I am using a discrete time multinomial logit with five competing risks, plus non-choice being the sixth outcome. It is pretty clear that I have to deal with the IIA issue somehow. The asmprobit command won't work as I cannot ascribe alternative-specific values in the periods in which no choice is made. So, I was wondering whether multinomial logit with random effects, as it is describe in the gllamm manual, can be of help here? In other words, does it help me relax the IIA assumption and to what extent? Many thanks!
Dimitrije