Markov switch state probabilities forecast

For technical questions regarding estimation of single equations, systems, VARs, Factor analysis and State Space Models in EViews. General econometric questions and advice should go in the Econometric Discussions forum.

Moderators: EViews Gareth, EViews Moderator

Nicotrader87
Posts: 13
Joined: Sun Jul 28, 2019 2:26 pm

Markov switch state probabilities forecast

Postby Nicotrader87 » Mon Jul 29, 2019 7:22 am

Hello everyone,

I need help estimating markov switch models. In particular I cannot understand how I can carry out forecast checks for probability states.

can you help me?
 

Nicotrader87
Posts: 13
Joined: Sun Jul 28, 2019 2:26 pm

Re: Markov switch state probabilities forecast

Postby Nicotrader87 » Tue Jul 30, 2019 4:21 am

can anyone help me?

EViews Glenn
EViews Developer
Posts: 2616
Joined: Wed Oct 15, 2008 9:17 am

Re: Markov switch state probabilities forecast

Postby EViews Glenn » Tue Jul 30, 2019 10:10 am

I'm afraid that I don't understand the question.

Nicotrader87
Posts: 13
Joined: Sun Jul 28, 2019 2:26 pm

Re: Markov switch state probabilities forecast

Postby Nicotrader87 » Tue Jul 30, 2019 11:05 am

Simply
I need to forecast a state probability but the software it allow the forecast of only fitting series.


EViews Glenn
EViews Developer
Posts: 2616
Joined: Wed Oct 15, 2008 9:17 am

Re: Markov switch state probabilities forecast

Postby EViews Glenn » Tue Jul 30, 2019 4:07 pm

I'm sorry, but that is correct.

Nicotrader87
Posts: 13
Joined: Sun Jul 28, 2019 2:26 pm

Re: Markov switch state probabilities forecast

Postby Nicotrader87 » Tue Jul 30, 2019 4:14 pm

sorry but I do not understand.
The markov model is born to return the probabilities of state, right?
so how can I carry out goodness tests of the probabilities without being able to carry out forecasts in sample and out of sample?

startz
Non-normality and collinearity are NOT problems!
Posts: 3497
Joined: Wed Sep 17, 2008 2:25 pm

Re: Markov switch state probabilities forecast

Postby startz » Tue Jul 30, 2019 4:21 pm

EViews produces several different measures of within-sample probabilities. Try Proc Make Regime Probabilities

I don't believe EViews does out-of-sample probability forecasts, but I don't think it should be hard to do.

Nicotrader87
Posts: 13
Joined: Sun Jul 28, 2019 2:26 pm

Re: Markov switch state probabilities forecast

Postby Nicotrader87 » Tue Jul 30, 2019 4:44 pm

sure, in fact I can extract the probabilities but how can I test the probabilities if in the "forecast" function I can only get the series fitted?

how do I test insample and outofsample without the forecast function?

startz
Non-normality and collinearity are NOT problems!
Posts: 3497
Joined: Wed Sep 17, 2008 2:25 pm

Re: Markov switch state probabilities forecast

Postby startz » Tue Jul 30, 2019 4:51 pm

Nicotrader87 wrote:sure, in fact I can extract the probabilities but how can I test the probabilities if in the "forecast" function I can only get the series fitted?

how do I test insample and outofsample without the forecast function?

Well, the extracted probabilities are in-sample forecasts.

If you want to forecast out of sample, I believe you just take the final state probabilities and run them out in a GENR statement multiplying by the estimated transition probabilities.

Nicotrader87
Posts: 13
Joined: Sun Jul 28, 2019 2:26 pm

Re: Markov switch state probabilities forecast

Postby Nicotrader87 » Tue Jul 30, 2019 4:57 pm

startz wrote:
Nicotrader87 wrote:sure, in fact I can extract the probabilities but how can I test the probabilities if in the "forecast" function I can only get the series fitted?

how do I test insample and outofsample without the forecast function?

Well, the extracted probabilities are in-sample forecasts.

If you want to forecast out of sample, I believe you just take the final state probabilities and run them out in a GENR statement multiplying by the estimated transition probabilities.



please for me it is very important can you make this process clearer?
maybe with a screenshot?

startz
Non-normality and collinearity are NOT problems!
Posts: 3497
Joined: Wed Sep 17, 2008 2:25 pm

Re: Markov switch state probabilities forecast

Postby startz » Tue Jul 30, 2019 5:52 pm

Have you tried Proc Make Regime Probabilities?

Nicotrader87
Posts: 13
Joined: Sun Jul 28, 2019 2:26 pm

Re: Markov switch state probabilities forecast

Postby Nicotrader87 » Wed Jul 31, 2019 1:19 am

startz wrote:Have you tried Proc Make Regime Probabilities?


obvious but how to make the forecast? with make probabilities get only the estimate in samples.

startz
Non-normality and collinearity are NOT problems!
Posts: 3497
Joined: Wed Sep 17, 2008 2:25 pm

Re: Markov switch state probabilities forecast

Postby startz » Wed Jul 31, 2019 6:08 am

If the last in sample probability is last_prob and the transition probabilities are p11 and p01, then I think you want something like

prob(T+1) = last_prob*p11 + (1-last_prob)*p01
prob(T+2) = prob(T+1)*p11 + (1-prob(T+1))*p01

etc

Nicotrader87
Posts: 13
Joined: Sun Jul 28, 2019 2:26 pm

Re: Markov switch state probabilities forecast

Postby Nicotrader87 » Wed Jul 31, 2019 7:58 am

Please i need the help with command line :oops:





https://ibb.co/1nRz1wd


Return to “Estimation”

Who is online

Users browsing this forum: No registered users and 2 guests