Long-run solution of the estimated STAR models

[Brc]

Hello,

I have an estimated nonlinear STAR model, which now behaves like a linear model. I would like to estimate its long-run solution. When the transition function takes the extreme values of F = 0 and F = 1, I want to determine where the model converges after applying a shock. I could not find a suitable code for this purpose. I believe an approach similar to an impulse–response function might be appropriate. What kind of code should I run?

Thank you,

Best.

[TomDoan]

Have you looked at TARMODELS.RPF: https://estima.com/webhelp/topics/tarmodelsrpf.html? Note (as that notes when analyzing the cycles in the branches) that the behavior of two branches tells you nothing about the overall behavior of the model. For instance, that has two non-stationary branches but the overall process is stationary since each branch drives the process towards the threshold to switch to the other branch.