Misspecified proportional hazard models and confirmatory analysis of survival data

Publisher Website
Google Scholar
Abstract
In survival analysis of follow‐up studies with staggered entry, it may be requested to examine a chance finding at an interim analysis in a confirmatory analysis based on ‘independent evidence’. Keiding et al. (1987) suggested incorporating residual lifetimes of initial survivors into the confirmatory analysis and indicated heuristically that this provides valid independent evidence, assuming that the statistical model is true. The present paper provides a full mathematical proof of a much more robust result, namely that the re‐use of original survivors provides an asymptotically independent confirmatory analysis even under a broad class of model misspecification, allowing unmeasured patient heterogeneity.

This publication has 0 references indexed in Scilit: