bivariate common cycle test a la Engle-Kozicki(1993)

[rylsnat]

Has anyone experimented with the bivariate common cycles tests as proposed in Engle-Kozicki (1993) and Engle-Vahid (1993) papers? I am having problem understanding the Engle-Kozicki procedure for the common feature test since it links estimation of the bivariate VAR(1) in conjunction with the 2SLS and LIML estimations. I would appreciate any help on this issue, including any sample code.

[TomDoan]

This would show the three ways to compute this for the pair jpngnp and gbrgdp. The test statistics are the LIML specification test for the LIML estimator and the J-statistics for the two directions of the IV estimates.

Code: Select all

instruments constant jpngnp{1} gbrgdp{1}
@liml jpngnp
# constant gbrgdp
linreg(inst) jpngnp
# constant gbrgdp
linreg(inst) gbrgdp
# constant jpngnp

[nightwalk]

Dear Tom,

I have isolated the series with significant ARCH effects and i want to estimate now the bivariate Engle and Kozicki common features test, which is performed (to the best of my knowledge) by two-stage least squares and the use of instrumental variables.

Can you give me such a code in RATS by explaining each stage? I think that the code you have provided in the forum is not the suitable one.

Warm regards