Some practical advice
may be given for regression modeling, based on the theoretical
considerations, as outlined before, and practical experience
in regression modeling from the authors and others. The practical
advice is split in two parts:
"What Not To Do"
might also be labeled "Pitfalls and Common Errors," based on
the experience of the authors in assessing the medical literature.
The advice relates especially to data sets of small to moderate
size, e.g. including less than 200 patients or less than 200
events, where model uncertainty and parameter uncertainty are
key problems.