Fluctuation of estimates in an EM procedure for categorical data
Seong Ho Kim, Sung-Ho Kim, Sung-Ho Kim, Sung-Ho Kim · Journal of Statistical Computation and Simulation · 2005
Estimates from an EM algorithm are somewhat sensitive to the initial values for the estimates, and this sensitivity is likely to increase when the model becomes larger and more complicated. In this paper, we examined how the estimates fluctuate during an EM procedure for a recursive model of categorical variables. It is found that the fluctuation takes place mostly during the initial stage of the procedure and that it can be reduced by applying a Bayes method of estimation. Both real and simulated data are used for illustration.