Event History Analysis

Event history analysis is a family of methods for modeling whether and when events occur over time while accounting for censored cases, including life tables, Kaplan-Meier estimation, and hazard regression models.

사건사 분석은 중도절단된 사례를 고려하면서 사건이 일어나는지와 언제 일어나는지를 모형화하는 방법군으로, 생명표, 카플란-마이어 추정, 위험률 회귀모형을 포함합니다.

How do I analyze how long it takes until an event happens when some cases have not experienced it yet?

일부 사례에서 아직 사건이 일어나지 않았을 때 사건 발생까지 걸리는 시간은 어떻게 분석하나요?

Event history analysis models the timing of events, such as job changes, marriages, or policy adoptions, using all cases including those whose event has not yet occurred by the end of observation. It fits data with a defined starting point, a recorded time, and an event indicator, and is the social science term for what biostatistics calls survival analysis.

When to use Event History Analysis

• Time until a life event, such as first job, marriage, or migration
• Observations censored because follow-up ended before the event occurred
• Comparing the timing of events across groups or periods

When to use something else

• Only whether the event happened within a fixed window matters and there is no censoring: use logistic regression
• Several mutually exclusive event types: use competing risks analysis

Assumptions and requirements

• Clearly defined time origin and event
• Non-informative censoring
• Independent cases
• Proportional hazards when Cox regression is used

What the analysis reports

• Survival or cumulative incidence curves
• Median time to event with confidence interval
• Hazard function or life table
• Log-rank test for group differences
• Hazard ratios from regression models

How to report it

The median time to [event] was [value] [time units] (95% CI [[value], [value]]), and the timing of [event] differed by [group], log-rank χ²([df]) = [value], p = [value].

Related methods

• Kaplan-Meier Estimator
• Cox Proportional Hazards Regression
• Discrete-Time Hazard Model

Run Event History Analysis on Quorant