Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Measuring the predictive ability of models


From   Paul Millar <paul.millar@shaw.ca>
To   statalist@hsphsun2.harvard.edu
Subject   st: Measuring the predictive ability of models
Date   Sat, 19 Mar 2005 16:28:50 -0700

Louis,

Another measure I like to use is the Proportional Reduction in Errors (PRE) provided by the model.  The nice thing about this measure is that it can be applied to all models and means (more or less) the same thing: a measure of the prediction provided by the model.  There is a routine to do this at:
www.ucalgary.ca/~pemillar/stata.htm

After installation, you type pre after the estimation command.
Example:
mlogit nomvar iv1 iv2
pre

For logistic/probit models, lstat does this (it provides more information but is also less clear), but this also works for ordinal and multinomial models.

- Paul

----- Original Message -----
From: louis boakye-yiadom <louisby@hotmail.com>
Date: Saturday, March 19, 2005 2:48 pm
Subject: Re: st: Predictive ability of probit model

> Thanks, Richard. What I'm looking for is the percentage of cases 
> correctly 
> predicted, as well as, the percentage incorrectly predicted. By 
> the way, I 
> just tried the lstat, but I got the following error message:
> 
> . lstat;
> last estimates not found
> r(301);
> 
> The lstat command just followed the "dprobit" command in my do 
> file. Thank 
> you.
> 
> Louis
> 
> 
> >From: Richard Williams <Richard.A.Williams.5@nd.edu>
> >>Dear All,
> >>After estimating a probit model, how can I let stata provide me 
> with the 
> >>predictive ability of the model? I suspect this is easy to do, 
> but I'm 
> >>fairly new to stata and I couldn't find the answer from the 
> online help 
> >>feature. Many thanks.
> >>
> >>Louis
> >
> >Louis, could you clarify what you mean by "predictive ability" of 
> a model?  
> >There is the -lstat- command, which will show you the percentage 
> of cases 
> >correctly classified.  The -predict- command can compute 
> predicted 
> >probabilities for each case, as well as residual statistics that 
> indicate 
> >problems in prediction.  The default output includes various 
> measures of 
> >model fit, including the model chi-square and McFadden's Pseudo 
> R^2; still 
> >more can be obtained with the -fitstat- command that is part of -
> spostado-. 
> >  The -adjust- command is great for plugging in 
> hypothetical/representative 
> >values and seeing what the predicted probabilities are under the 
> estimated 
> >model (e.g. low income males, high income females.) If you can be 
> more 
> >specific about what you want, we might be able to point you in a 
> specific 
> >direction.
> 
> 
> *
> *   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/



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