Title

Fitting Event-History Models to Uneventful Data

Author(s)/Creator(s)

Douglas A. Wolf, Syracuse University. Center for Policy ResearchFollow
Thomas M. Gill, Yale School of Medicine ; Dorothy Adler Geriatric Assessment CenterFollow

Document Type

Working Paper

Date

12-2007

Keywords

Disability, semi-Markov process, duration analysis

Language

English

Series

Working Papers Series

Disciplines

Econometrics

Description/Abstract

Data with which to study disability dynamics usually take the form of successive current-status measures of disability rather than a record of events or spell durations. One recent paper presented a semi-Markov model of disability dynamics in which spell durations were inferred from sequences of current-status measures taken at 12-month intervals. In that analysis, it was assumed that no unobserved disablement transitions occurred between annual interviews. We use data from a longitudinal survey in which participants' disability was measured at monthly intervals, and simulate the survival curves for remaining disabled that would be obtained with 1- and 12-month follow-up intervals. The median length of an episode of disability based on the 12-month interval data is over 22 months, while the "true" median, based on the 1-month interval data, is only one month.

Additional Information

Working paper no. 101

Harvest from RePEc at http://repec.org

Recommended Citation

Wolf, Douglas A. and Gill, Thomas M., "Fitting Event-History Models to Uneventful Data" (2007). Center for Policy Research. 65.
https://surface.syr.edu/cpr/65

Source

Metadata from RePEc

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.