Applications of the grade of membership technique to event history analysis: Extensions to multivariate unobserved heterogeneity

Analyses of the event histories of social and service utilization processes are often difficult because of a lack of adequate theory to specify the distributional form of any latent heterogeneity [J. Heckman and B. Singer, The identifiability of the proportional hazards model. Rev. Econ. Studies 51, 231-241 (1984); ibid., A method for minimizing the impact of distributional assumptions in econometric models for duration data. Econometrics 52, 271-320 (1984).] or the form of the basic hazard rate [J. Trussel and C. Hammerslough, A hazards-model analysis of the covariates of infant and child mortality in Sri Lanka. Demography 20, 1-26 (1983).]. In this study we present an analytic strategy that deals with both questions nonparametrically using a conditional likelihood approach. The model is illustrated using 24 months of follow-up data on Supplemental Security Income beneficiaries in Type D (Nursing Home) living arrangements. The parameter estimates can be used in standard life table computations to determine the amount of time expected to be spent in different residential and payment statuses for different analytically identified classes of beneficiaries. © 1986.

Duke Scholars

Author Patrick J. Eric Stallard Social Science Research Institute

Author Kenneth G. Manton Demographic Studies

Author Anatoli I. Yashin Social Science Research Institute