Journal of Official Statistics, Vol.7, No.3, 1991. pp. 261–293
A Multivariate Event History Model Based Upon Fuzzy States: Estimation From Longitudinal Surveys With Informative Nonresponse
Kenneth G. Manton, Eric Stallard, and Max A. Woodbury
Abstract:Event history analysis of health and functional changes at advanced ages is made especially difficult by interactions of the events of interest and the temporal structure of the survey's observational plan. In particular, because of the rapid pace and high prevalence of morbid events at advanced ages, downwardly biased estimates of the rates of health and functional status changes may be produced by censoring occurring within the field survey period. We illustrate such problems in a large recent U.S. national longitudinal survey of chronically disabled elderly persons and present statistical procedures to deal with this type of censoring. Such bias may exist in many different survey designs when applied to very elderly populations. However, to adequately adjust for the longitudinal nonresponse bias, the design must provide detailed information on individual health transitions and mortality. Collection of the necessary information may require the use of list sampling where vital status follow-up is nearly complete.
Keywords:Event history analysis; censoring; response bias; longitudinal surveys; disability; population aging.
Copyright © Statistics Sweden 1996-2018. Open AccessISSN 0282-423XCreated and Maintained by OKS Group