The RATS Software Forum

Hello,

I am working on cointegration analysis and want to do some statistical test which allow me to choose the best option between various options for intercept and trend. Specifically, I am interested in doing G(r) statistic, something that was mentioned in “A closer look at the U.S. Housing Market: Modeling Relationships among Regions”, Yunus & Swanson, Real Estate Economics, 2013, V41: pp 542-568 at page 551. This paper mentioned utilizing G(r) statistics from “The role of the constant and linear terms in cointegration analysis of nonstationary variables”, Soren Johansen, Econometric Reviews Vol. 13 , Iss. 2, pp 205-209, 1994. I read this paper too, it mentions five general models 2.4 to 2.8 on page 208.

I have read CATS manual section 1.4 and “The cointegrated VAR model”, Katrina Juselius chapter 6 which talks about five cases but there is no procedure available to test this automatically.

Is there a way that some kind of code can be written in RATS (or CATS) which can make choice among five options?

Thanks

victor

Isn't that just a likelihood ratio test (with a non-standard distribution) where you vary (only) the DET term. For instance, for testing DET=CONSTANT (general drifts) with DET=RTREND (trend in the cointegrating relations).