Flexible parametric survival models

Estimate the shape of the hazard, and report something a patient could use.

Taught by

Sara Ekberg, PhD

Associate Director of Applied Biostatistics

Over 15 years of experience in register-based epidemiological research, and a lecturer with a passion for teaching statistics to non-statisticians.

Research cancer survival · survivorship · register-based research · teaching

PhD in Medical Sciences, Karolinska Institutet, 2021

Register your interest

What you will be able to do

• Say what a Cox model leaves unestimated, and what having a baseline hazard buys you
• Fit a Royston-Parmar model, choose the flexibility of its baseline, and defend the choice
• Model a time-dependent effect, and read a hazard ratio that changes with time
• Produce standardised survival curves and absolute-scale contrasts, and report them

Before you start

• Take first: Survival analysis

Curriculum

12 steps: 11 lectures and 1 practical.

Provisional — this course is still being written, and the outline will change.

Week 1 · why the baseline hazard is worth having

• 01WatchWhy the baseline hazard is worth modelling22 min
• 02WatchFrom a constant rate to a smooth one26 min
• 03PracticalPractical · splitting the timescale, and the shape it shows45 min

Week 2 · the model, and its splines

Opens 7 days after you start.

• 04WatchThe Royston-Parmar model26 min
• 05WatchRestricted cubic splines, and what the coefficients are not24 min
• 06WatchHow many knots, and where24 min

Week 3 · time-dependent effects

Opens 14 days after you start.

• 07WatchProportional hazards, and how to see it fail24 min
• 08WatchFitting a time-dependent effect26 min
• 09WatchThe hazard ratio as a function of time26 min

Week 4 · predictions worth reporting

Opens 21 days after you start.

• 10WatchPredictions for a covariate pattern26 min
• 11WatchStandardised survival curves28 min
• 12ReadContrasts on the absolute scale, and how to report them22 min