Statalist


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

Re: Re: st: Cluster/PCA for predicting a binary outcome?


From   n j cox <n.j.cox@durham.ac.uk>
To   statalist@hsphsun2.harvard.edu
Subject   Re: Re: st: Cluster/PCA for predicting a binary outcome?
Date   Wed, 18 Jul 2007 18:11:52 +0100

Frank Harrell has a characteristically incisive summary
of logistic vs discriminant analysis in his "Regression
modeling strategies" (Springer, New York, 2001), with
references. The burden is: go with logistic.

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

(By the way, he ain't fond of stepwise.)

E. Paul Wileyto

You're looking for discriminant analysis (look up Stata help on
discrim)...  It's like PCA, but you can think of it as selecting the
loadings to maximize the F-value if you did a 1-way ANOVA by your class
variable.

But, for a binary outcome, many simply use logistic regression.

K Jensen wrote:

> I am trying to predict a binary outcome from a set of correlated
> variables.  Rather than using logisitic regression to include the
> variables one-by-one in the model, I was wondering if there was a way
> in Stata of generating sets or functions of the variables?  Maybe
> something similar to PCA, except that you would be trying to explain
> the variation in the binary outcome rather than the whole dataset.

*
*   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