The RATS Software Forum

Dear Tom,

I am dealing with a stationary process with structural breaks( in means) and say T=200 . I want to know if it is possible, lets say if you are at t=30 and you want to know what the mean of the series is going to be for next 7 periods.what sort of method should I be using or will be appropriate. Is there any way of predecting breaks.

Best wishes,
Rosen

A MS model would predict the probability of a break. It doesn't really predict the time of a break in any particularly interesting way---if you're in regime 1, the probability of a switch to regime 2 in the next few periods will follow a geometric distribution. Now that's a model with regimes that switch back and forth, while it sounds like you have different types of breaks, which would require rather heroic assumptions to predict out of sample.