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

st: RE: PSEUDO R2


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: PSEUDO R2
Date   Fri, 5 Dec 2003 15:09:06 -0000

Two comments: 

1. Squared correlation between observed and predicted 
is one measure you can use here. There are other measures
called pseudo-R^2 which don't all necessarily agree numerically 
or algebraically. There is some discussion at 

http://www.stata.com/support/faqs/stat/rsquared.html

and since then there has been at least one 
thread on Statalist. The matter arises quite 
often on Statalist, so an archive search may be 
fruitful. 

2. If you have missing data on any of the 
covariates, then the model will not fitted
on the corresponding observations even if 
the response is non-missing in those observations. 
In that circumstance -predict- will be predicting 
out of sample. It's safer to stipulate -if e(sample)-. 
This issue is touched upon in the FAQ above. 

Nick 
n.j.cox@durham.ac.uk 

Antonio Rodrigues Andres
 
> I am estimating grouped data logistic models using glogit or simply
> using regress with analytical weights
> 
> generate double lcrp=log(ntprop/(pop-ntprop))
> generate double wt1=ntprop*(pop-ntprop)/pop
> 
> 
> glogit ntprop pop clprop $xvars  $zvars $tvars  
> regress lcrp clprop $xvars  $zvars $tvars [w=wt1]
> 
> If I want to get a measure of goodness of fit I might use 
> the squared
> correlation between y and yhat 
> 
> regress lcrp clprop $xvars $zvars $tvars 
> predict yhat1
> corr yhat1 lcrp
> 
> Is that correct?

*
*   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