flexsurv: flexible parametric survival modelling in R. Supplementary examples

This vignette of examples supplements the main flexsurv user guide. Keywords:˜survival. 1. Examples of custom distributions 1.1. Proportional hazards generalized gamma model Crowther and Lambert (2013) discuss using the stgenreg Stata package to construct a pro-portional hazards parameterisation of the three-parameter generalised gamma distribution. A similar trick can be used in flexsurv. A four-parameter custom distribution is created by defining its hazard (and cumulative hazard) functions. These are obtained by multiplying the built-in functions hgengamma and Hgengamma by an extra dummy parameter, which is used as the location parameter of the new distribution. The intercept of this parameter is fixed at 1 when calling flexsurvreg, so that the new model is no more complex than the generalized gamma AFT model fs3, but covariate effects on the dummy parameter are now interpreted as hazard ratios.> library(flexsurv)> hgengammaPH <- function(x, dummy, mu=0, sigma=1, Q){ + dummy * hgengamma(x=x, mu=mu, sigma=sigma, Q=Q