I'm looking for advice on planning the development of a prognostic model. Does anyone have a reference giving advice on the types of studies required for the various stages of development? I have the Pepe book but this is aimed more at diagnostic tests and it's hard to translate to what is needed for a prognostic model.
We have a relatively small case-control study on which the initial exploratory work has been carried out on. We would like to take four models forward as potentially promising but I am proposing a follow-on study (also retrospective case-control) to use ROC curves to identify optimal cutpoints for risk stratification and get better estimates for the weights of each factor in the model. Is it sufficient at this stage to for example use 10 cases per factor in the model and match with the same number of control subjects? Or could we calculate sample size based on achieving minimum predictive values at this point? Would we determine the best model at this stage and then proceed to a prospective validation study or is there another step first?
Many thanks in advance for any suggestions,
Thanks
Kim
You may leave the list at any time by sending the command
SIGNOFF allstat
to listserv@..., leaving the subject line blank.
