Modelling survival data using flexible parametric models in Stata using stpm3: concepts and modelling choices.

This course will cover the modelling of survival (time-to- event) data using flexible parametric survival models in Stata.

The course will make use of the stpm3 command (released on SSC in June 2023) which has many advantages over its predecessor, the stpm2 command (released in 2008/9).

The course will cover general modelling issues that are useful when using survival models for either description, prediction or understanding causality.

The course is aimed at individuals who have an understanding of standard survival analysis methods (e.g. censoring / Kaplan-Meier curves / Cox proportional hazards models).

The course will cover the following topics:

• The advantages (and a few disadvantages) of using flexible parametric survival models.
• Choosing the number and location of knots to model the effect of time.
• Modelling non-linear effects of covariates (using splines and other functions).
• Choice of scale: Log cumulative hazard, log hazard and other scales.
• Relaxing the proportional hazard assumption.
• Predictions of survival, hazard and other useful functions.
• Making contrasts between covariate groups.
• The use of marginal predictions (regression standardization) and contrasts to quantify the effect of exposures/treatment.
• The use of marginal predictions to assess model fit and predictive performance.
• How to perform a sensible sensitivity analysis.
• How to avoid model convergence problems.