East Coast Ballroom
Unified method for Markov chain transition model estimation using incomplete survey data (306777)
*Duncan Ermini Leaf, USC Schaeffer Center for Health Policy and EconomicsKeywords: EM algorithm, Markov chain, microsimulation, missing data, Monte Carlo
The Future Elderly Model and related microsimulations are modeled as Markov chains. Their transition models are estimated from longitudinal survey data, such as the Health and Retirement Study, Panel Study of Income Dynamics, and similarly structured international surveys. The use of surveys for estimation poses several incomplete data problems, including coarse and irregular spacing of interviews, data collection from subsamples, and structural changes to data over time. A general method based on the Expectation-Maximization algorithm is presented for estimating transition models using incomplete survey data. The method is demonstrated on data from the Health and Retirement Study.