MM-05 Survival and duration analysis (Kaplan-Meier estimator; Nelson-Aalen estimator; Cox proportional hazards; time-to-event analysis)
Estimate the time until an event while using records whose event has not yet been seen, and produce survival curves with intervals and a median duration.
What it does
Estimates the time until an event while using records whose event has not yet been observed. Artefact: a survival curve with intervals and a median duration. Read curve separation and censoring counts.
When to use it
Use when you need to know how long something lasts or how fast an event happens and some cases are still ongoing.
When it does not fit
Needs durations with an event flag. Misfit: heavy censoring, or non-proportional hazards for Cox models (not confirmed).
Result format
The result has a survival curve with intervals, a median duration, and the counts of records with an observed event and of censored records. It states whether the curves for the compared groups separate.
Requirements on the material
| Property | Requirement type | Requirement |
|---|---|---|
| Structure | Applies in some cases | requires=structured |
| Content type | Always applies | requires=tabular |
| Unit of analysis | Always applies | requires=record |
| Variable types | Raises a warning | prefers=ratio; prefers=nominal |
| Time structure | Applies in some cases | requires=panel; requires=cohort; requires=continuous stream |
| Level of detail available | Always applies | requires=record-level |
| Appraises source credibility | Applies in some cases | no |
Not restricted on: How the sample was chosen, Number of sources, Whether sources can be joined, Amount of material, Known data problems, Producer interest, Source dependence, Origin, Document genre, Language, Personal information, Security classification, Licence.
| Minimum material | at least 30 records with a start date, an end or last-seen date and an event flag |
|---|---|
| Additional sources needed | none |
| Not for (rules) | none |
| Not for (text) | Wrong choice when there is no event flag or when most records are censored. Cox models also assume proportional hazards, which was not confirmed in the source. |
Choosing and ordering
| Kind of technique | Primary technique, runs in step 5.3 |
|---|---|
| Phase tag | analyse |
| Other phase tags | none |
| Kind of finding | trend; difference |
| Question answered | how much; what next |
| Strongest statement it supports | show a link between things |
| Needs as input |
|
| Makes as output |
|
| Must exist first | None |
| Loop target | return to verify and prepare |
Library checks by stakes
Each stakes level includes the checks of the level below it. Checks that need a person apply at high stakes only, so the process leaves them out at lower stakes (decision 38). Repeatability: deterministic.
| Stakes | Check | Person involved | Counts as a different method |
|---|---|---|---|
| exploratory | input validation | none | no |
| decision-informing | input validation | none | no |
| decision-informing | code execution for arithmetic | none | no |
| decision-informing | sensitivity analysis | none | no |
| high-stakes | input validation | none | no |
| high-stakes | code execution for arithmetic | none | no |
| high-stakes | sensitivity analysis | none | no |
| high-stakes | assumption and limitation statement | none | no |
| high-stakes | second technique with different assumptions | none | conditional |
Failure modes and limits
| Model suitability | model plus code |
|---|---|
| Typical model failure | Ignores censoring and averages the observed durations. |
| Error sources | input data; statistical assumptions |
| Certainty limits | Certainty drops with heavy censoring or few events (imprecision), when censoring is related to the outcome (risk of bias) and when hazards are not proportional for Cox models (inconsistency). |
Related techniques
- should follow PE-05 Missing data diagnosis and treatment (missingness analysis; imputation; complete case analysis; non-response weighting): missing end dates must be separated from censored records
- pairs with CK-13 Multiple-testing correction (Benjamini-Hochberg false discovery rate control; Bonferroni; Holm): comparing many subgroup curves needs multiple-testing correction
Evidence and tag basis
Documented. The only source read is a software package page. It names the estimators and shows an event-observed flag, but it does not define censoring. (lifelines on PyPI)
Tag basis: input profile inference; selection inference; sequencing inference; robustness inference. All tag values are inference unless a source is named.