We compared various modeling strategies for logistic regression models in a paper in Stat Med 2000
and Medical Decision Making 2001.
The code for application of alternative strategies to sample5 is here (S-plus).
We proposed an updating procedure towards recalibrated values for a previously developed model, including shrinkage to avoid overoptimistic predictions. This idea builds on previous work by Hans van Houwelingen, see e.g. his
Stat Med 2000 paper. Simulations for our Stat med 2004 paper were done with S-plus code.