Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: out-of-sample predictions with GLM


From   "Maarten Buis" <M.Buis@fsw.vu.nl>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: out-of-sample predictions with GLM
Date   Thu, 26 Jul 2007 14:44:42 +0200

--- jc1926@gmx.de wrote:
>  Using proportions data, I am estimating a GLM of 
> the type described in Papke and Wooldridge's 1996 
> article entitled "Econometric methods for fractional 
> response variables...", Journal of Applied 
> Econometrics, vol 11, 619-632.
> 
> The Stata command to estimate the model is:
> glm y x1 x2, fam(bin) link(logit) robust

> One nice feature of the model is that the in-sample 
> predictions (logically) sum to one. 

Do you mean that:
the predicted proportion for observation 1 + 
the predicted proportion for observation 2 +
the predicted proportion for observation 3 +
.
.
the predicted proportion for observation _N = 1 ? 

This is not true. I also cannot think of any situation 
in which that would be desirable. 

A situation where you want proportions to add up to 1 
within an observation occurs when you have for each 
observation multiple proportions, e.g. for each city 
you have the proportions of the city budget spent on 
safety, education, transport, etc. In this case the 
proportions within a city should add up to 1. A model 
that deals with this situation is -dirifit-,see: 
http://home.fsw.vu.nl/m.buis/software/dirifit.html  

Hope this helps,
Maarten

-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology 
Vrije Universiteit Amsterdam 
Boelelaan 1081 
1081 HV Amsterdam 
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434 

+31 20 5986715

http://home.fsw.vu.nl/m.buis/
-----------------------------------------


*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index