Dear Statalist users,
Is there a way to estimate multinomial outcome with a binary endogenous variable?
My outcome variables are several measures of health status in binary form (bad/good). In addition, I want to address attrition due to death by including death as an outcome
I tried using -gsem- in Stata 13.1 but I am having the same coeff and se as with independent multinomial logit and logit specifications (I also tried different covstructure options).
I also tried user written -cmp-, but having difficulty in getting the ml algorithm to converge.
I still believe -gsem- is the way to go and I am wondering whether anyone can guide me to materials which discusses the proper specification of covariance structures in this case.
Thanks,
dhiman
Is there a way to estimate multinomial outcome with a binary endogenous variable?
My outcome variables are several measures of health status in binary form (bad/good). In addition, I want to address attrition due to death by including death as an outcome
I tried using -gsem- in Stata 13.1 but I am having the same coeff and se as with independent multinomial logit and logit specifications (I also tried different covstructure options).
I also tried user written -cmp-, but having difficulty in getting the ml algorithm to converge.
I still believe -gsem- is the way to go and I am wondering whether anyone can guide me to materials which discusses the proper specification of covariance structures in this case.
Thanks,
dhiman
