Advanced Statistical Modelling (Full-time)

Module code: MD7444

Survival analysis

Survival analysis is concerned with data where we measure the time to an event. In medicine we are often interested in death, i.e. we want to keep people alive and make them live longer. We thus measure the time (from some suitable starting point) to death. The outcome of interest is the time to an event and we are unlikely to observe the event on all subjects.

Topics covered may include:

• Introduction to survival analysis – censoring, the survival function, the Kaplan-Meier estimate, median survival, life-tables
• Comparing survival curves - the log rank test, hazard ratio
• Modelling survival - proportional hazards, exponential model, Weibull model, the Cox model, partial likelihood, tied observations, interpreting regression coefficients
• Model fitting and selection
• Model checking and regression diagnostics for the Cox model

Multilevel modelling

You will be introduced to multilevel modelling for the analysis of hierarchical and repeated measures data for both continuous and binary outcomes. You will have practical experience of working with a range of software packages and will also have the opportunity to critique articles published in the medical literature using these techniques.

Topics covered will include:

• Introduction to multilevel (hierarchical) data structures
• Summary methods
• Introduction to multilevel modelling for hierarchical data
• Multilevel models for longitudinal/repeated measures data
• Multilevel models using classical and Bayesian approaches
• Repeated binary data analysis
• Alternative analysis: Generalised estimating equations (GEEs)
• Bayesian approaches to multilevel modelling.