Similar to ths question I would like to know how to create a calibration curve without binning my predictions. What makes my situation different, is that I'm using icenReg for my interval-censored outcome (proportional hazards weibull model for interval censoring). I don't mind using predictions at the right-end of the intervals. The rms package doesn't work with these models. No time varying covariates (icenReg doesn't accept these- instead i had to make the assumption, which is likely true anyway, that observations are independent, despite multiple observations coming from same patients).
The only way I can think of visualising my model's calibration is to predict at right end of the interval (when outcome is known), bin cases by probability {r} cut(probability, breaks = seq(0, 1, by = 0.01) and calculate crude incidence of the event in each group. Not ideal- but what else can I do?
On a side note: Is obtaining the kaplan-meier estimate at the right end of the interval any better?
