About multilevel truncated models

[johnsoncat]

Greetings,

I would appreciate your advice on the following. There are the commands 'tpossion' and 'tnbreg' in Stata to implement the single-level truncated Poisson model and the single-level truncated negative binomial model. (However, as far as I know, we cannot do these models in MLwin directly. ) So, how can I implement a multilevel truncated Poisson model or a multilevel truncated negative binomial model in runmlwin?
What's more, just the same question in implementing multilevel zero-inflated models.
Thanks.

[GeorgeLeckie]

Hi John,

In terms of count models, you can only do multilevel poisson and negative binomial models in MLwiN.

You can't do truncated or zero-inflated versions of these models.

What you could attempt is to specify your model a multilevel poisson model and then request MLwiN to generate the the winbugs model statement and assocated files for you.

You could then tweak the winbugs model statement to incorporate the truncated or zero-inflated part. You will have to explore the WinBUGS manuals to work out how to do this.

http://www.mrc-bsu.cam.ac.uk/bugs/winbu ... ents.shtml

You can use the winbugs suite of Stata commands to run your modified model in WinBUGS and to pull the results back.

http://www2.le.ac.uk/departments/health ... s_main.pdf

You can also use the mcmcsum which we have written (bundled with runmlwin) which can also analyse the MCMC chains pulled back from WinBUGS.

Best wishes

George

[johnsoncat]

[quote="GeorgeLeckie"]

Hi George,

Thank you for all your assistance. I truly appreciate your help in resolving the problem. ^-^