Advanced Survival Analysis: When the Cox Model Isn't Enough

The Cox model is the most useful tool in survival analysis and also the most over-extended. It assumes one event, one transition, one measurement occasion, independent subjects, exact event times, a hazard ratio that stays put, and a question that was never causal to begin with. Real data violates all seven assumptions before lunch. This course is about what to do next. Ten chapters, each anchored on a failure mode of the standard toolkit and the established method that repairs it: estimands and restricted mean survival time when hazards are not proportional; flexible parametric and cure models when you must extrapolate past your follow-up; multi-state models when subjects move between more than two states; recurrent-event models when the event happens again and again; frailty models when subjects are clustered or heterogeneous; interval-censoring and truncation methods when the sampling scheme quietly rigged your risk sets; joint models when the covariate is a noisy biomarker measured every few months; Bayesian survival models when events are scarce and external information is not; causal survival analysis when the question is what would happen if we treated them rather than who is at risk; and machine learning for survival, evaluated honestly enough to discover how often it fails to beat a penalized Cox model. Every chapter states the estimand before the estimator, derives the key identity rather than asserting it, and ends with a decision rule for practice. Code is given in both R and Python wherever both are genuinely usable, and where Python has no mature implementation — multi-state, frailty, and joint models — the course says so plainly instead of faking it with a wrapper nobody uses.

  • Level: advanced
  • Duration: 10 weeks
  • 11 sections
  • Price: $99.99

Topics: advanced survival analysis, non-proportional hazards, restricted mean survival time, RMST, estimand framework, weighted log-rank, MaxCombo, flexible parametric survival, Royston-Parmar splines, relative survival, cure models, multi-state models, competing risks, Fine-Gray, cause-specific hazard, Aalen-Johansen, illness-death model, recurrent events, Andersen-Gill, PWP, mean cumulative function, frailty models, shared frailty, unobserved heterogeneity, non-collapsibility, interval censoring, Turnbull estimator, left truncation, length-biased sampling, joint models, longitudinal biomarkers, dynamic prediction, Bayesian survival analysis, piecewise exponential, causal inference, marginal structural models, inverse probability weighting, g-formula, target trial emulation, causal survival forests, random survival forests, gradient boosted survival, DeepSurv, DeepHit, discrete-time survival, SHAP, R, Python, flexsurv, mstate, JMbayes2, lifelines, scikit-survival, pycox

What's inside

  1. Advanced Survival Analysis: When the Cox Model Isn't Enough
  2. Chapter 1: Life After Proportional Hazards
  3. Chapter 2: Flexible Parametric Models, Extrapolation, and Cure Fractions
  4. Chapter 3: Multi-State Models
  5. Chapter 4: Recurrent Events
  6. Chapter 5: Frailty, Clustering, and the Heterogeneity You Cannot See
  7. Chapter 6: Interval Censoring, Truncation, and the Sampling You Did Not Notice
  8. Chapter 7: Joint Models for Longitudinal Biomarkers and Survival
  9. Chapter 8: Bayesian Survival Analysis
  10. Chapter 9: Causal Survival Analysis
  11. Chapter 10: Machine Learning for Survival, and an Honest Benchmark

Read a free sample chapter