Markov switch regression Forecast
Posted: Wed Mar 04, 2020 9:58 am
Hi everyone I have a problem in interpreting the output of a markov switch model with Eviews.
When we estimate such a model we get two outputs, one is the fitted series and one is the probability states.
Suppose we have a sample 1-20000 values, if I estimate a model in sample I get probabilities (2 state) and if I go to get the forecast (static) of the whole sample (1-20000), I get a fitted series.
Now my doubts are:
Do I use the probability or the Fit to determine the model's performance?
Also, why do the probability and the fit differ from each other?
When we estimate such a model we get two outputs, one is the fitted series and one is the probability states.
Suppose we have a sample 1-20000 values, if I estimate a model in sample I get probabilities (2 state) and if I go to get the forecast (static) of the whole sample (1-20000), I get a fitted series.
Now my doubts are:
Do I use the probability or the Fit to determine the model's performance?
Also, why do the probability and the fit differ from each other?