Chauvet (1998) Dynamic Factor with Markov Switching

[ftggog]

Hi,

I'm very new to this, so forgive me for my basic questions.

I want to estimate a dynamic factor with Markov switching equation (see http://faculty.ucr.edu/~chauvet/ier.pdf).

Question is to see if an economy might enter recession soon.

Any idea how I can do this in Eviews? I'm using Eviews 9.5.

Thanks v much in advance.

[trubador]

Such hybrid models are quite complex and difficult to implement. Even if you manage to handle it, the computational burden would be high. I would try dynamic factor and Markov switching models separately to see if there is any evidence to support the combination of the two.

[ftggog]

Thank you.

Would you be able to show me how to do dynamic factor models in Eviews?

I was able to do a very simple Markov switching model to gauge recession probability, but I'll try to supplement it with others.

[ftggog]

I've also read somewhere that dynamic factor models are a specific variation of a state space model?

If that's right, what's peculiar about the dynamic factor model.

Thanks so much in advance.

[trubador]

Yes, that is correct. It is because dynamic factor(s) can be treated as latent variable(s) and can be put into state space form. You can search the forum for examples in EViews: http://forums.eviews.com/search.php?st= ... mic+factor

[ftggog]

Thanks v much!

[ftggog]

Hi, one more question.

How do I download the raw data from a regime probability graph (I'm using the switching regression).

Right now, I can view the graph but I can't find a way to download the data

[trubador]

http://www.eviews.com/help/EViews%209%2 ... l#ww190518

[ftggog]

Thank you. A few more questions if I may on Markov switching models. I'm reading through the Eviews manual at the same time.

- Can we add additional independent/explanatory variables to the switching model?
- How should we interpret the Log(Sigma) coefficient and its statistical significance? (in both a constant parameter and time varying parameter model)
- How do we interpret the transition matrix parameters? (in both constant parameter and time varying parameter model)
- How do we determine the lags for AR terms, and how should we interpret its statistical significance?
- How do we gauge if the model is robust and stable?

Apologies, very basic questions over here.

[trubador]

Unfortunately, these questions are related to very fundamental knowledge of Markov Switching modeling, not to implementation of it in EViews. You are better off referring to a textbook for more details.