[R] cross-validation / sensitivity anaylsis for logistic regression model

[Cody_Hamilton at Edwards.com]

Dylan,

You might like the validate() function in the Design library.  It validates
several model indeces (e.g. R^2) using resampling.  There is some
discussion on this function (as well as on validating your model via
resampling) in the book on S programming by Carlos Alzola and Frank Harrell
(available at
http://biostat.mc.vanderbilt.edu/twiki/pub/Main/RS/sintro.pdf).

Regards,
   -Cody
                                                                           
                                                                           
                                                                           
                                                                           
                                                                           




                                                                           
             Dylan Beaudette                                               
             <dylan.beaudette@                                             
             gmail.com>                                                 To 
             Sent by:                  r-help at stat.math.ethz.ch            
             r-help-bounces at st                                          cc 
             at.math.ethz.ch                                               
                                                                   Subject 
                                       [R] cross-validation / sensitivity  
             05/14/2007 04:38          anaylsis for logistic regression    
             PM                        model                               
                                                                           
                                                                           
             Please respond to                                             
             dylan.beaudette at g                                             
                 mail.com                                                  
                                                                           
                                                                           




Hi,

I have developed a logistic regression model in the form of (factor_1~
numeric
+ factor_2) and  would like to perform a cross-validation or some similar
form of sensitivity analysis on this model.

using cv.glm() from the boot package:

# dataframe from which model was built in 'z'
# model is called 'm_geo.lrm'

# as suggested in the man page for a binomial model:
cost <- function(r, pi=0) mean(abs(r-pi)>0.5)
cv.10.err <- cv.glm(z, m_geo.lrm, cost, K=10)$delta

I get the following:
cv.10.err
    1     1
0.275 0.281

Am I correct in interpreting that this is the mean estimated error
percentage
for this specified model, after 10 runs of the cross-validation?

any tips on understanding the output from cv.glm() would be greatly
appreciated. I am mostly looking to perform a sensitivity analysis with
this
model and dataset - perhaps there are other methods?

thanks

--
Dylan Beaudette
Soils and Biogeochemistry Graduate Group
University of California at Davis
530.754.7341

______________________________________________
R-help at stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.