[R] multinomial probit

[Kosuke Imai]

Dear Charlie,
  The multinomial probit model estimates the covariance matrix of the
latent variables in addition to the coefficents. In your case with 5
alternatives, there are 9 parameters in the covariance matrix. This means
that even if you only have intercepts in your model (no covariate), you
will have a total of 13 parameters to estimate with only 28 data points.  
If you still want to fit the model, you need an informative prior. The
default prior is quite diffuse, and so you need to adjust p.var, p.df, and
p.scale arguments in mnp(). See the full documentation available at
http://www.princeton.edu/~kimai/research/MNP.html and the paper for more
details and examples:

Imai, Kosuke, and David A. van  Dyk. (2005). ``A Bayesian Analysis of the  
Multinomial Probit Model Using Marginal Data Augmentation.''  Journal of 
Econometrics, Vol. 124, No. 2 (February), pp. 311-334.

Hope this helps,
Kosuke

---------------------------------------------------------
Kosuke Imai               Office: Corwin Hall 041
Assistant Professor       Phone: 609-258-6601 
Department of Politics    eFax:  973-556-1929
Princeton University      Email: kimai at Princeton.Edu
Princeton, NJ 08544-1012  http://www.princeton.edu/~kimai
---------------------------------------------------------

> From: Charlie Sprenger <c.sprenger at gmail.com>
> Date: December 3, 2004 4:09:19 AM CST
> To: r-help at stat.math.ethz.ch
> Subject: [R] multinomial probit
> Reply-To: Charlie Sprenger <c.sprenger at gmail.com>
>
> Hello All,
>
>  I'm trying to run a multinomial probit on a dataset with 28 data
> points and five levels (0,1,2,3,4) in the latent choice involving
> response variable.
>
> I downloaded the latest mnp package to run the regression. It starts
> the calculation and then crashes the rpogram. I wish I could give the
> error message but it literally shuts down R without a warning.
>
> I'm using the R that's been ported to OS X. Is it possible that there
> are RAM restrictions and that when something computationally heavy
> goes through the program is cut off?
>
> As well, is there actually a "probit" method for polr from nnet 
> package?
>
> All help appreciated as masters thesis topics are due soon.
>
> Charlie Sprenger
> UCL - student
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! 
> http://www.R-project.org/posting-guide.html
>

--
Andrew D. Martin, Ph.D.
Department of Political Science
Washington University
Campus Box 1063
One Brookings Drive
St. Louis, MO 63130
(314) 935-5863 (Office)
(314) 753-8377 (Cell)
(314) 935-5856 (Fax)

Office: Eliot Hall 326
Email: admartin at wustl.edu
WWW:   http://adm.wustl.edu