Predicted probabilities after multinomial logit

[jmflavel]

Hello,
I ran a 2 level multinomial logit model using panel data (repeated measures over time on a number of individuals). I ran the model using MQL1 first to obtain starting values then used MCMC. I am modelling form of employment and I used the following model (abbreviated version posted here, actual model contained more controls).

runmlwin foe cons unemp_lag ptime_lag postgfc female kid04 kid514 edu1 edu2 edu3,level2(id:cons) level1(tpid) discrete(distribution(multinomial) link(mlogit) denominator(cons) basecategory(2)) mcmc(on) initsprevious nopause

I want to estimate predicted probabilities from this model. I am aware of how to do this using gllamm (via gllapred) but have not been able to determine how to do it using runmlwin.

I came across the following posts in my research into this

http://www.cmm.bristol.ac.uk/forum/view ... ?f=3&t=199
http://www.cmm.bristol.ac.uk/forum/view ... ?f=3&t=486

but neither of these covers multinomial models. Can anybody please advise how I might extend this to a multinomial logit model to obtain predicted probabilities?

Regards,
Joanne

[GeorgeLeckie]

Chapter 10 of the MLwiN user manual contains an example...

http://www.bristol.ac.uk/cmm/media/runm ... sponses.do


Best wishes

George

[jmflavel]

Thanks so much George. I did look at the manual but I somehow missed that.