On Friday 28 November 2008 (18:53:43), Anthony A. Pezzola wrote:
> Hi all,
>
>
>
> I was wondering is anyone could suggest the best option for estimating
> multilevel ordered logit (probit) models. My options seem to be running
> MIXOR or BUGS through R, using glamm via STATA, or running the models in
> HLM, which I do not own. Is MCMC rally the best way to go?
Anthony:
From my limited experience, there is no single best way to fit these kinds of
models. It depends on the complexity of the model, your resources, and your
patience.
MIXOR and glamm use quadrature approaches, which are highly accurate but can
be very slow if the model is very complex. (MIXOR does afaik only very simple
models.)
HLM uses several versions of Laplace approximation, which are less accurate
(which is less an issue if the clusters are large) but much faster.
MCMC approaches are very flexible, but they make take long until convergence
and convergence is sometimes difficult to assess. Also you have to make
decisions about prior distributions for the parameters of your model.
So if your model is relativly simple (a two-level model, that is, only one
level of random effects) MIXOR may be your choice.
If you have money, but less patience and large clusters you might use HLM.
Otherwise your choice is either glamm or bugs. But your mileage may vary.
HTH
Martin
--
-------------------------------------------------
Dr. Martin Elff
Department of Social Sciences
University of Mannheim
A5, Room 328
68131 Mannheim
Germany
Phone: ++49-621-181-2093
Fax: ++49-621-181-2099
E-Mail: [log in to unmask]
Homepage:
http://webrum.uni-mannheim.de/sowi/elff
http://www.sowi.uni-mannheim.de/lspwivs/
-------------------------------------------------
**********************************************************
Political Methodology E-Mail List
Editors: Melanie Goodrich, <[log in to unmask]>
Delia Bailey, <[log in to unmask]>
**********************************************************
Send messages to [log in to unmask]
To join the list, cancel your subscription, or modify
your subscription settings visit:
http://polmeth.wustl.edu/polmeth.php
**********************************************************
|