Elimination of variables in the STR/Threshold model

[Estef M.]

Good afternoon
I have a question regarding threshold models. After including all variables in the specification, in case you find that a variable it not significant, how can you eliminate it? I am talking about eliminating variables that are not significant in the linear (nonlinear) part but significant in the nonlinear (linear) one.
Thank you beforehand.

[EViews Rebecca]

Eliminate in what sense?

[Estef M.]

In the sense that if you initially introduce all possible variables (and lags) and you find that some of them are not statistically significant, you can take as the next step the estimation of the model without those non-significant variables. It is quite typical in order to smooth or "refine" the model.
Thank you beforehand.