Confidence inteefals foe teamsitiom probabilities in too-state markov chains

Abstract

This paper develops approximations for the distributions of the maximum likelihood estimators of the transition probabilities in a finite state, irreducible, aperiodic Markov chain* These distributions are then used to compute confidence intervals for the transition probabilities, and to make sample size projections. The techniques used are presented by first considering the two-state Markov chain, and then the many state Markov chain is discussed.

Keywords:

• Estimation theory/ approximationsg sampling distributions