Discrete Measurement of Time and Interval Censoring in Event History Analysis

37 Pages Posted: 6 Jun 2015

Date Written: June 4, 2015

Abstract

A problem in event history analysis is that time is measured imprecisely. Events are typically known to occur within discrete time units (e.g. day, month or year). Discrete measurement of the start and end time of an event leads to a known interval within which the event duration falls. The event duration is interval censored. When ignored, interval censoring is shown to introduce considerable bias to parameter estimates and heighten the risk of inference errors. I show that treating the duration as an interval reduces bias and improves the performance of hypothesis tests. Replications of analyses from four political science articles in leading journals demonstrate that substantive inferences depend on the use of appropriate methods for interval censored duration data. I also develop a software package that can be used to estimate the Cox proportional hazards model with interval censoring.

Suggested Citation

Desmarais, Bruce A., Discrete Measurement of Time and Interval Censoring in Event History Analysis (June 4, 2015). Available at SSRN: https://ssrn.com/abstract=2614922 or http://dx.doi.org/10.2139/ssrn.2614922

Bruce A. Desmarais (Contact Author)

Pennsylvania State University ( email )

University Park, State College, PA 16801
United States

HOME PAGE: http://sites.psu.edu/desmaraisgroup

Register to save articles to
your library

Register

Paper statistics

Downloads
25
Abstract Views
272
PlumX Metrics