Time-varying transition probabilities in MS VAR

[renis]

Dear Tom,

I have been trying to estimate a three-regime MS-VAR (and regime-dependent impulse-response functions) with time-varying transition probabilities. The time-varying regime probabilities are a function of macroeconomic variables. I am wondering whether time-varying transition probabilities are feasible in the code (eev_mcmc.rpf) that follows the Ehrmann-Ellison-Valla(2003) approach and that uses mssysregression.src source file. If so, how those macroeconomic variables can be input and how the code can be modified?

Many thanks indeed.

Kind regards.

[TomDoan]

1. Do you have a reference for that? Gibbs sampling for time-varying transition probabilities would require sampling a multinomial logit.

2. Is there a reason that you want to estimate such a complicated model? Did you successfully estimate a three regime model without time-varying transitions?