Markov Switching model

[jspro eur]

I was wondering whether it is possible to see which observations are allocated to the different states when a markov switching model is used. Does anybody know a command I could use?

[Behram Wali]

I have the same question regarding mswitch in Stata 14.1. How to see which values of response/outcome are assigned to different states. E.g. if my dependent variable is a continuous variable consisting of positive, negative and zero values, and i run a two state mswitch, is there any way to see that to which state zeros are assigned?

Thank you

[Ashish Rajbhandari (StataCorp)]

Dear jspro and Behram,

In a Markov-switching model, we do not know with certainty the assignment of observations to states. Rather we can only estimate the probability of an observation being in a given state.

You may use the -predict- command with option -pr- after you fit a model with -mswitch- to obtain the probabilities.

Best,
Ashish

[River Huang]

I fit a three-regime model, and wish to obtain the filtered probabilities for those three states. How can I do that?

[River Huang]

Dear jspro and Behram,

In a Markov-switching model, we do not know with certainty the assignment of observations to states. Rather we can only estimate the probability of an observation being in a given state.

You may use the -predict- command with option -pr- after you fit a model with -mswitch- to obtain the probabilities.

Best,
Ashish

I fit a three-regime model, and wish to obtain the filtered probabilities for those three states. How can I do that? Thanks in advance.

[Behram Wali]

Hi River, After you run the model, you can get filtered probabilities by typing:

predict fprob, pr smethod(filter)

.

For details, please look at http://www.stata.com/manuals14/tsmsw...postestimation

[River Huang]

Hi River, After you run the model, you can get filtered probabilities by typing:

predict fprob, pr smethod(filter)

.

For details, please look at http://www.stata.com/manuals14/tsmsw...postestimation

I did that before. But, shouldn't we obtain "three" filtered probabilities for each of the three states. Your suggested command, as I did earlier, only provided one series of probability.

[Catherine Bloom]

Dear Professor Huang,

I think you may try the following command:

predict pr_31 pr_32 pr_33, pr smethod(filter)

I hope this helps.

Best wishes,
Catherine