Scott,
As I think somebody mentioned before, the GLLAMM package in Stata
(www.gllamm.org) should be able to estimate multilevel logit models.
Section 9.3 of the gllamm manual (which can be found at
http://www.bepress.com/cgi/viewcontent.cgi?article=1160&context=ucbbiostat)
has a discussion of a random intercept multinomial model. Gllamm uses the
Stata syntax for generalized linear models, where both the distribution
family of the dependent variable and the model link function are specified.
In your case, the multinomial logit model (with random intercepts) would go
something like:
gllamm y x1 x2 x3, i(countryid) family(binom) link(mlogit) base(basecat)
weight(wt)
where basecat is the arbitrary base category and wt is the variable that
indexes the frequency weight (if the data have been collapsed which is the
usual case) for the level-1 units.
While gllamm cannot yet estimate the unrestricted multilevel multinomial
probit model, one advantage of using gllamm is that you can also specify one
alternative familiy related to the multionomial probit: scaled ordinal
probit --family(soprobit)-- which assummes that the depedent variable can be
ordered although the order of the categories has to estimated from the data
(this model is thus less restrictive that the mixed ordinal
probit --family(binom) link(oprobit)-- but more parsimonous than the mixed
multnonimial logit).
hope this helps,
Omar.
----- Original Message -----
From: "Orsolya K. Lazar" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Friday, December 02, 2005 9:19 PM
Subject: Re: [POLMETH] Hierarchical MNP
> According to my reading of the manual, HLM6 (Raudenbush, Bryk, Cheon,
> Congdon &
> du Toit) can do hierarchical multinomial LOGIT. Raudenbush and Bryk's book
> (Hierarchical Lineral Models: Application and Data Analysis Methods)
> also has a
> discussion of hierarchical multinomial models.
>
> --
> Orsolya K. Lazar
> Dept. of Political Science
> University of Arizona
> Tucson, AZ 85721-0027
> [log in to unmask]
>
>
> Quoting "Scott D. McClurg" <[log in to unmask]>:
>
>> I'm working on a project where I'm looking at a multinomial dependent
>> variable and where individuals are nested in states. So, there is both a
>> hierarchical process and a multinomial dependent variable. Is anyone
>> familiar with the use of a multilevel, multinomial probit model in
>> political
>> science? If so, could you please direct me to any (accessible) technical
>> papers and applications?
>>
>> If I get enough varied responses, I'll summarize them and post them to
>> the
>> list.
>>
>> Thanks in advance!
>>
>> Best,
>>
>> scott
>>
>> ____________________________________________
>> Scott D. McClurg
>> Assistant Professor
>> Department of Political Science
>> Southern Illinois University
>> Mailcode 4501
>> Carbondale, IL 62901-4501
>> Phone: 618/453-3191
>> FAX: 453-3163
>>
>> E-mail: [log in to unmask]
>> Internet: http://www.siu.edu/~mcclurg
>>
>> **********************************************************
>> Political Methodology E-Mail List
>> Editor: Karen Long Jusko <[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
>>
>> **********************************************************
>
> **********************************************************
> Political Methodology E-Mail List
> Editor: Karen Long Jusko <[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
>
> **********************************************************
**********************************************************
Political Methodology E-Mail List
Editor: Karen Long Jusko <[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
**********************************************************
|
