Survival analysis
Survival analysis is a statistical approach to modeling the time until an event occurs while accounting for incomplete or censored event-time observations.
Effective Data Analysis
Survival analysis is a statistical approach for analyzing the timing of events while accounting for observations whose events have not yet occurred.
- How it is used
- Frames time-to-event questions in clinical, operational, subscription, and user-conversion contexts.
- How deep it goes
- Defines time-to-event prediction, survival curves, and Kaplan–Meier analysis but does not develop model equations or assumptions.
- Examples
- Uses patient survival, machine failure, customer subscription duration, subscription drop-off, and treatment-control survival curves.
- What you can do after
- Apply it via Python's lifelines library to estimate survival curves and compare event timing across groups.
Read how Effective Data Analysis explains survival analysis →
Financial AI in Practice
A statistical approach to analyzing the time until an event occurs while accounting for observations whose event time is unknown or incomplete.
- How it is used
- Tracks when individual loans default to support portfolio-risk monitoring and underwriting decisions.
- How deep it goes
- Works through synthetic loan durations, event indicators, and a Kaplan–Meier curve, but does not develop a borrower-covariate model.
- Examples
- Uses 50 synthetic loans, roughly 10% defaults within 12 months, and a stepwise Kaplan–Meier survival curve.
- What you can do after
- Apply it via lifelines to plot and interpret default timing for loans with observed and unobserved events.
Read how Financial AI in Practice explains survival analysis →
Related concepts
- vintage analysiscontrasts with · Vintage analysis evaluates credit performance by grouping accounts according to their origination period and tracking outcomes across subsequent periods.
- Kaplan–Meier estimatoruses · The Kaplan–Meier estimator is a nonparametric estimator of a survival function that accounts for right-censored observations.
- covariateuses · A covariate is an observed or supplied variable included in a statistical, causal, or predictive model to describe, explain, predict, or adjust for variation in an outcome.
- time-based analysisused by · Time-based analysis models how outcomes and risks evolve over an observation period rather than only at a fixed endpoint.
- right censoringaddresses · Right censoring occurs when an observation is known to have remained event-free up to the end of observation but its eventual event time is unknown.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.