NA values in Markov models

[evasileiou]

Hello!!

I try to run a Markov model taking as a sample 2,472 observations.

When I run the model using 2 regimes I have all the Information that the EVIEWS provides available.

However, when I try to examine if more than 2-regimes(states) models perform better that the 2_regimes, I have not the Information regarding the Std.Error, z-Statistic and the Prob.

e.g. Schwarz criterion suggests that the 3regimes model is better than the 2-regimes, but without the aforementioned Information I cannot report it.

Does anyone know if there is an explanation why this happens or/and how I could find a solution?

Thanks in advance.

[EViews Glenn]

Probably a singularity. It is often difficult to estimate Markov models with multiple regimes. Sometimes there simply isn't enough information to estimate a third (or even second) regime. You can try different starting values using the randomization tools. You might also want to post your workfile.

[evasileiou]

I tried to attach one of the files, but there is message "The extension wf1 is not allowed." How could I attach a wf1 file?

[EViews Gareth]

email it to support@eviews.com