Parameter Estimation in Logistic Regression for Transition, Reverse Transition and Repeated Transition from Repeated Outcomes

Covariate dependent Markov models dealing with estimation of transition probabilities for higher orders appear to be restricted because of over-parameterization. An improvement of the previous methods for handling runs of events by expressing the conditional probabilities in terms of the transition probabilities generated from Markovian assumptions was proposed using Chapman-Kolmogorov equations. Parameter estimation of that model needs extensive pre-processing and computations to prepare data before using available statistical softwares. A computer program developed using SAS/IML to estimate parameters of the model are demonstrated, with application to Health and Retirement Survey (HRS) data from USA.