Re: logistic regression and oversampling

We came across this issue the other day. We have case-control data
that an investigator suggested be "re-weighted" back to the population
proportions. However, I came across the reference below that says if
you are analysing by logistic regression, there is no need to reweight
as long as there is an intercept term and you include the strata used
for oversampling as covariates in the model. This is also the same
argument for using the odds ratio as an estimate of relative risk for
case-control studies.

Prentice and Pyke (1979), Logistic disease incidence models and
case-control studies, Biometrika, 66:403-411.

.