# Calibration curve for pooled logistic regression

Suppose I am fitting a logistic regression to the outcome cbind(successes, failures) rather than to the outcome y \in (0,1) (i.e., I’m doing pooled logistic regression) to account for differential exposure within an observation period. How do I draw smoothed calibration curves for the fitted model? Do I plot the smoothed sum of fractional periods as a function of predicted probabilities rather than the smoothed sum of events?