The RATS Software Forum

Hi,

I took the Switching Models and Structural Breaks course, and have many great codes on Markov Switching (thank you!!).
I tried to take a crack at it to modify the SWARCH model or MS ARCH model so that i can do the switching on the mean equation's coefficients, instead of what the examples are doing on the variance equation only.
i.e.
y = a + b1x1+b2x2+u
where u = sqrt(h)v
v is iid with mean 0 and variance of 1
h~ f(lag terms of u^2)

and focus on switching of b1 and b2

Would anyone point me to the right direction, as i am a little lost.

THANK YOU!

This log likelihood would be similar to that of the SWARCH - it's just that the squared residuals for computing the variances and the current residual for computing the final likelihood would be regime-dependent rather than the variances.