4

I'm surveying algorithms to estimate the transition probabilities of a discrete Markov chain and I found that EM approach could used. However I am not able to find any simple description on how to implement the EM algorithm to fit a discrete Markov chain giving a data sequence. Is there any R function that does this?

Nick Cox
  • 48,377
  • 8
  • 110
  • 156
Giorgio Spedicato
  • 3,444
  • 4
  • 29
  • 39
  • I can help you if you can explain the problem a bit better. Since you wish to use EM, there should be an observed (known) component and an unobserved (hidden) component. In your case, what is observed and what is latent/hidden? The mention of EM and Markov chains collectively bring to mind the Baum-Welch algorithm for Hidden Markov Models. If that is the case, there is a package called HMM in CRAN. – Zhubarb Dec 02 '13 at 16:30

2 Answers2

1

you should check "Estimation of the transition matrix of a discrete-time Markov chain" by Bruce A. Craig. This article describe how to do what you want.

wish this helps

Marco A. Aquino
  • 11
  • 1
1

Before you proceed with any algorithms for hidden Markov models please read one of the keystone papers in HMMs:

http://www.cs.ubc.ca/~murphyk/Bayes/rabiner.pdf

In it you will find a wonderful overview of HMMs as well as 3 key algorithms to solve the 3 problems in HMMs that are usually of interest!

Also there are several implementations of HMMs in R:

Here is one http://cran.r-project.org/web/packages/HiddenMarkov/HiddenMarkov.pdf

bdeonovic
  • 8,507
  • 1
  • 24
  • 49