st: Re: Multinomial with Conditioning Variable unique to each response?

["R.E. De Hoyos" <redeho@hotmail.com>]

Javier,

I couldn't quite follow your question. In the multinomial logit model what 
varies across outcomes (in your case years) is the parameter not the 
independent variables, hence you do have a unique set of explanatory 
variables across alternatives for every unit of observation.

You said:
> More specifically, say I want to model the probability of an event

> occurring in any of four years, 1982, 1983, 1984 and 1985.
This sounds more like a conditional logit (-help clogit-). Although the 
conditional and the multinomial logit are mathematically the same, the 
nature of the problem they pretend to capture is completely different. In 
the conditional logit you are estimating the probability of occurrence of 
ANY event given a set of explanatory variables. In this case, the parameters 
are the same across outcomes and what varies is the explanatory variables. 
In the multinomial logit what you estimate is the probability of observing a 
particular event given a set of explanatory variables (given a normalization 
needed for identification, the results are interpreted as relative 
probabilities compared to the base outcome.)

I hope this helps,

Rafa
________________________
R.E. De Hoyos
Faculty of Economics
University of Cambridge
CB3 9DE, UK
www.econ.cam.ac.uk/phd/red29/

----- Original Message ----- 
From: "Javier Miranda" <javier.miranda@census.gov>
To: <statalist@hsphsun2.harvard.edu>
Sent: Friday, July 29, 2005 3:48 PM
Subject: st: Multinomial with Conditioning Variable unique to each response?



> I want to estimate a Multinomial Logit Model where one of the conditioning 
> variables is unique across alternatives.
> More specifically, say I want to model the probability of an event 
> occurring in any of four years, 1982, 1983, 1984 and 1985.
> Now, say one of the conditioning variables is unque across alternatives; 
> for example, assume the interest rate differs each year.
> Now, if I estimate this model and generate the predicted probabilities I 
> end up with a set of probibilites for each interest rate by year.
> This is not right. Any suggestions as to what model I should be looking 
> at?
>
> Thanks Much
> --javier miranda
>
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/