Con dence Intervals for Markovian Models

This paper introduces a new multinomial approach unifying the computation of confidence intervals for Markovian models. Starting from a method used for homogeneous Markov chains, we show that it can be applied on models incorporating a hidden component. We consider three models derived from the basic homogeneous Markov chain: the Mixture Transition Distribution (MTD) model, the Hidden Markov Model (HMM), and the Double Chain Markov Model (DCMM). Compared to existing methods, our proposal can be used on data sets of any size without requiring extensive computing.