Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: aic bic adjusted R-squared with svy command


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: aic bic adjusted R-squared with svy command
Date   Fri, 19 Feb 2010 19:21:17 -0000

The tacit reference can be imputed here as 

Harrell, Frank E. 2001. Regression modeling strategies. New York: Springer. 

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

Lachenbruch, Peter

Harrell has a lot more to say, not much good.  Some recent work by Austin and Tu (I think) in the American Statistician(?) or Statistics in Medicine (?) has suggested doing stepwise on bootstrap samples and using this in lieu of a stepwise.  It removes some of the bias.
You might also look up Frank's book.  I can't find it on my shelves, so I must have lent it to someone.

Ronan Conroy

On 18 Feabh 2010, at 07:22, Björn Bünger wrote:

> For each estimation type I want to conduct a (manual) stepwise  
> regression and examine AIC / BIC or adjusted R-squared on each step.

There is general agreement that adjusted R-squared for stepwise models  
underestimates the model shrinkage. Harrell says that you ought to  
adjust R-squared for the total number of variable considered in the  
model, not the number actually used in the final model.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   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   |   Site index