any parameter instability for nonlinear least square (nlls)?

[mjchang36]

Hello everyone:
I wolud be very pleased if you inform me whether there is any written code for testing parameter stability for nonlinear least square model in RATS or suggest me any applicable test for that. Thanks with regards.
mjchang36

[TomDoan]

If you have a known break point, you can just do a sample split Chow test. If you have an unknown break point, you can use @FLUX or @APGRADIENT to look for a general break. Those are both designed to take the scores directly off an instruction estimating by maximum likelihood, so you have to do an extra step with NLLS of multiplying the derivatives by the residuals, such as

Code: Select all

nlls(frml=nlconst,derives=scores)
do i=1,%nreg
   set scores(i) = scores(i)*%resids
end do i
@flux
# scores
@apgradienttest
# scores

[mjchang36]

thank you.