Event History Analysis

Historical Development

Event history analysis is also referred to as duration, survival, or reliability analysis, depending on the substantive origins of the discussion (medicine and engineering for the latter two terms, respectively). Early applications involved life table analysis by Kaplan and Meier (1958), but the historical roots can be traced back even further to the late 1600s (Hald, 1990). There was an increased use of the technique during World War II because of concerns over the expected reliability of military equipment. The path-breaking work of D. R. Cox (1972) is credited with another period of expansion in the use of event history analysis as a result of his development of semiparametric techniques. His work is likely to be heralded as one of the top statistical achievements in the 20th century. Applications in medicine and the social sciences have increased greatly as a result of the less restrictive semiparametric Cox regression model and its various extensions, which are built upon the mathematics of counting processes (Therneau & Grambsch, 2000).

Parametric and Semiparametric Models

Both parametric and semiparametric models are available in event history analysis. Analysts studying mechanical systems typically use parametric models, which assume that the time until the event of interest follows a specific distribution, such as the exponential. Studies of human behavior and biology typically use the less restrictive semiparametric Cox model, which leaves the particular distributional form of the duration times unspecified. Blossfeld and Rohwer (2002, pp. 180, 263) argued that social science theory rarely provides the justification for a specific parametric distribution and instead advocated for use of the Cox model.

...