Piece-wise exponential (Additive Mixed) Modelling Tools for Survival Analysis

Abstract

The piecewise exponential model provides one possible framework for the abstraction of the estimation task from the specific computing engine such that any of the survival problems mentioned above can be generalized into three steps: (1) data transformation, (2) model estimation using any software or package that supports the optimization of the Poisson log-likelihood (3) post-processing (e.g., prediction and visualisation). In this framework, left-truncation, time-varying effects and competing risks can be realised by appropriate data transformation and/or inclusion of interaction effects. No additional adjustments are required for the computation engine in step (2).