Linda wrote:

> Hi there!
>
> I wonder whether there are any good rule of thumb in choosing Markov states?
>
> Cheers,
> Linda

The answer to this also depends on your choice of estimation procedure. I assume
you are estimating it as in Rabiner's paper, which implies a maximum likelihood
estimation framework. In this case, calculate the penalized likelihood value
(e.g. using Akaike's model order selection criterion or, better still, Minimum
Description Length) for different number of states (state space dimension) and
pick the minimum.

Iead


.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to