Testing nonlinearity with robustness to outliers

[Estefaniame]

Good afternoon

I have the RATS code for doing linearity tests vs STAR models in presence of outliers ("RobustPoly" procedure). Nevertheless, I would like to apply it to a STR model. I do not find an easy way to do it. Is there already any command for it?

Thank you beforehand,

Estefanía

[TomDoan]

The same calculation can be done using @RegStrTest with the WEIGHT option rather than @STARTEST, with the iterated weighted least squares over the general linear regression.

[Estefaniame]

Good evening

The program works well with STR models! But I have a question, as the original program was for STAR models, the outliers were only present in variable Y. If I do the application to STR, do I have to define the outliers for each variable considered?

Thank you beforehand.

[TomDoan]

Outliers are defined as data points with unusual residuals. The "X" variables can have any sort of behavior.

[Estefaniame]

So I do only have to care for Y (dependent variable) regarding outliers? I can do the REGTEST for different thresholds afterwards.

Thank you!

[TomDoan]

They're outliers in the regression of interest, not outliers in the dependent variable.