12.3 Review: which method for which question
Table 2 collects the methods of the course by the question they answer. Three habits run through all of them: account for censoring rather than discarding it; state what the time origin and the event are; and check the assumptions of a model, proportional hazards above all, before reporting its single summary numbers.
| Question | Method | In R |
| What fraction survive to \(t\)? | Kaplan–Meier, Greenwood (Section 4) | survfit |
| What is the cumulative risk? | Nelson–Aalen, life tables (Section 5) | survfit |
| A smooth model, extrapolation | Weibull and others (Section 3) | survreg |
| Do groups differ? | log-rank tests (Section 6) | survdiff |
| Covariate effects on the hazard | Cox model (Sections 7 and 8) | coxph |
| Is the hazard ratio constant? | Schoenfeld residuals (Section 9) | cox.zph |
| Effects on time; predictions | AFT models (Section 10) | survreg |
| Several kinds of event | incidence, Fine–Gray (Section 11) | survfit, finegray |
| Clusters, repeated events | frailty, Andersen–Gill (Section 11) | frailty, cluster |
| How many patients? | Schoenfeld’s formula (Section 12) | simulation |
Questions on this section
Stuck on something here? Ask below and it stays attached to this topic.