The RATS Software Forum

Dear RATS users
I am looking for a code to estimate nonlinear (generalized) impulse response function in threshold error correction model (TECM).
I tried to it using basic code of Terasvirta (1994) in RATS discussion forum, but it is little differenct case.
So it would be great if you help me to code it.

Thank you for your time and best regards,

Jun

This is from the structural breaks and switching course. It's quite a bit more complicated than the calculation of the GIRF for the univariate case.

Corrected program posted 1 April 2011

Program file:


Data file:

Dear Tom

Thank you for your comment and codes. I have two questions about your code.
In your code, you created 5 GIRFs.
But I don't understand the relationship between numbers and created GIRFs.
For example, while GIRF(1) represents "GIRF for Spread to one S.D. shocks in Discount rate",
I don't understand why 1 means Spread's GIRF to shock in Discount rate, etc.
Another question is that is it possible to get (or split) GIRF for both positive and negative shock in discount rate?

Thank you for your consideration

j_lee wrote:Dear Tom

Thank you for your comment and codes. I have two questions about your code.
In your code, you created 5 GIRFs.
But I don't understand the relationship between numbers and created GIRFs.
For example, while GIRF(1) represents "GIRF for Spread to one S.D. shocks in Discount rate",
I don't understand why 1 means Spread's GIRF to shock in Discount rate, etc.
Another question is that is it possible to get (or split) GIRF for both positive and negative shock in discount rate?

Thank you for your consideration

Sorry. One of the series was mapped incorrectly. I've posted a corrected version above.

While it's possible to collapse the model down to fewer variables, it's much easier to leave everything in this form, with the differences, the levels and the error correction term as the five variables.

TomDoan wrote:This is from the structural breaks and switching course. It's quite a bit more complicated than the calculation of the GIRF for the univariate case.

Corrected program posted 1 April 2011

Program file:

sms_5_2.rpf

Data file:

irates.xls

it still uses the MC method not bootstrap method
am i right ? Tomn

iloverats wrote:it still uses the MC method not bootstrap method
am i right ? Tomn

It is randomizing over shocks, but not over coefficients. Those are being treated as fixed at the sample estimates.