Key points are not available for this paper at this time.
We present analyses for survival data obtained from a central cancer registry with passive follow-up. This method of data collection has the potential to produce unknown random losses which would affect estimates of survival. We show that non-parametrically estimated conditional distributions remove any effect of these unknown losses and that a compound mixture model estimates their magnitude. Lung cancer data are used to illustrate the procedures.
Tallis et al. (Fri,) studied this question.