[R] model diagnostics for logistic regression

[Frank E Harrell Jr]

Dylan Beaudette wrote:
> Greetings,
> 
> I am using both the lrm() {Design} and glm( , family=binomial()) to perform a 
> a logisitic regression in R. Apart from the typical summary() methods, what 
> other methods of diagnosing logistic regression models does R provide? i.e. 
> plotting an 'lm' object, etc. 

The best way of checking fit in a lrm is to do directed regression tests 
(e.g. extend linear effects into regression splines; interactions).  An 
omnibus test is in residuals.lrm as is partial residual plots.

> 
> Secondly, is there any facility to calculate the R^{2)_{L} as suggested by 
> Menard in "Applied Logistic Regression Analysis" (2002) ?

Don't know that paper but lrm has various indexes including 
Nagelkerke-Maddala R^2.

Frank

> 
> Any thought would be greatly appreciated.
> 
> Cheers,
> 
> 


-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University