Sample size for binary logistic prediction models: Beyond events per variable criteria

@topepos @BenVanCalster Uncertainty in the predictions has motivated a lot of the criteria on minimal sample size work in this area, e.g. https://t.co/QpMUjqu63T

@drg1985 Unfortunately, this is not true for calibration and discrimination. I tried to visualize this in one of the earlier papers: https://t.co/pGFTTKpzGW https://t.co/Z0N11QFJ6z

I am currently reading "Sample size for binary logistic prediction models..." from @MaartenvSmeden et al. and can't figure out which performance metrics were evaluated on a hold-out test data set (all?). Does anybody here know this? https://t.co/2nfs3N…

@noah_greifer One more caveat: if estimating the intercept matters to you, e.g. if you are developing a prediction model (https://t.co/QpMUjqu63T) or you do some marginalisation (https://t.co/3upbKigvEX) you need to re-adjust the intercept. This is auto…