R help - May 2012 - Interpreting output from the val.prob function in the rms package

Dear mail-list,

I have used the val.prob function from the rms package to validate a logistic
regression model. It is however not clear to me how to interpret the two
different curves that are generated (logistic vs. nonparametric). The
documentation doesn't enter into any detail on this point, and to a
non-statistician like myself it isn't obvious.

If I compare the output to that of the val.prob.ci function by E. Steyerberg, it
looks like the non-parametric curve is a smoothed line between the means of
quantiles. But what is then the logistic calibration line?

Most grateful for any help to understand this.

Best, Gustav


Gustav Nilsonne, MD, PhD
Postdoc
+46 (0) 736-798 743

Stockholm University
Stress Research Institute
106 91 Stockholm
gustav.nilsonne@stressforskning.su.se

Karolinska Institutet
Department of Clinical Neuroscience and Osher Center for Integrative Medicine
Retzius väg 8 A2:3
171 77 Stockholm
gustav.nilsonne@ki.se

	[[alternative HTML version deleted]]