[R] Package for .632 (and .632+) bootstrap and the cross-validation of ROC Parameters

Frank E Harrell Jr f.harrell at vanderbilt.edu
Thu Jul 12 19:22:24 CEST 2007


spime wrote:
> 
> Hi users,
> 
> I need to calculate .632 (and .632+) bootstrap and the cross-validation of
> area under curve (AUC) to compare my models. Is there any package for the
> same. I know about 'ipred' and using it i can calculate misclassification
> errors. 
> 
> Please help. It's urgent. 

See the validate* functions in the Design package.

Note that some simulations (see http://biostat.mc.vanderbilt.edu/rms) 
indicate that the advantages of .632 and .632+ over the ordinary 
bootstrap are highly dependent on the choice of the accuracy measure 
being validated.  The bootstrap variants seem to have advantages mainly 
if an improper, inefficient, discontinuous scoring rule such as the 
percent classified correct is used.

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



More information about the R-help mailing list