ABSTRACT
Model checking is an essential part of the modelling process, and there are a number of aspects of the fit of a model that need to be studied. These include the assumed functional form of explanatory variables in a model, whether there are survival times that are greater than might have been anticipated from the values of explanatory variables for certain individuals, and whether there are individuals with values of explanatory variables that have an undue influence on estimated hazard ratios or the risk score. Many model checking procedures are based on residuals, and various types, including Cox-Snell, martingale, deviance, Schoenfeld and score residuals, are reviewed. The way in which these can be used to examine various aspects of the fit of a model are then described and illustrated, with an emphasis on graphical techniques. The chapter concludes with a description of methods for checking the validity of the proportional hazards assumption, using both graphical methods and formal tests.
