Test for no remaining nonlinearity(additive test) and test for parameter consistency (Terasvirta nonlinear STAR model)

[Brc]

Hello, I want to ask RATs codes or RATs procedure files (.src) in order to do additive test against nonlinearity.


Firstly, I estimate nonlinear STAR model (by Terasvirta) (ESTAR model or LSTAR model). And then, I want to do additive nonlinearity test in order to detect additive transition function or different transition variable. I want to test for no remaining nonlinearity. For that, which test or which procedure should I do?

Similarly, after estimating nonlinear STAR model (by Terasvirta) (ESTAR model or LSTAR model), I want to test for parameter consistency. For that, which test or which procedure should I do by using RATS?


Thank you for helps.

[TomDoan]

See viewtopic.php?p=17011#p17011

[Brc]

Hello,

I am studying with univariate nonlinear model (by Terasvirta), STAR model. After estimating LSTAR model, I need to do test for no remaining nonlinearity. For that I will do Testing the adequacy of smooth transition autoregressive models (Eitrheima and Terasvirta, 1996) (Eitrheim, Ø., & Teräsvirta, T. (1996). Testing the adequacy of smooth transition autoregressive models. Journal of econometrics, 74(1), 59-75.)


Can you please show how to write down the codes of Additive Test for logistic transition function (LSTAR) by hand (manually)?

Thank you.

[TomDoan]

You don't have to do any of that. The test in the STARDIAGNOSTICS.RPF example works for both ESTAR and LSTAR---it lets the NLLS instruction do the work of taking the derivatives.

[Brc]

This is a different code from STARDIAGNOSTICS.RPF. I have another manual code for additive nonlinearity test. I want to use this manual code. But, I have this manual code for ESTAR, not for LSTAR. How can I write down these parts (TP, LP for LSTAR model)

You don't have to do any of that. The test in the STARDIAGNOSTICS.RPF example works for both ESTAR and LSTAR---it lets the NLLS instruction do the work of taking the derivatives.

[TomDoan]

I have no idea what any of that is and, aside from being really bad coding (no one should ever hard code the numbers into those formulas), it doesn't even look like it's a doing a correct implementation of the test for remaining threshold non-linearity. What is in the stardiagnostics.rpf program is a correct implementation of the test as well as being considerably simpler.

[Brc]

Okay thank you. I have one more question. How can I test for parameter consistency for LSTAR and ESTAR models?

I have no idea what any of that is and, aside from being really bad coding (no one should ever hard code the numbers into those formulas), it doesn't even look like it's a doing a correct implementation of the test for remaining threshold non-linearity. What is in the stardiagnostics.rpf program is a correct implementation of the test as well as being considerably simpler.

[TomDoan]

Use @FLUX. I added that to the end of the posted STARDIAGNOSTICS.RPF.