Page 1 of 1

about AIC/SC in Johansen cointegration test table

Posted: Sun Jul 22, 2012 2:51 am
by tino
Hi. I am not sure whether this is a bug, but allow me to ask.
I use the command "coint" to get a summary of Johansen cointegration test. As a part of its output, it produces the tables of log likelihoods, AIC and Schwarz criteria for different cointegration ranks and deterministic trend specifications.
Since I was curious about what "number of parameters" EViews uses when it calculates AIC and SC, I calculated the "number of parameters" from the AIC and the log likelihood of these tables, using a formula: #para=0.5*AIC*T+logL where T is the number of observations.
I am afraid that these numbers do not correspond to the number of parameters (with standard errors and t-stats) in the VECM. I use EViews version 7.2.
Thank you very much for your attention.

Re: about AIC/SC in Johansen cointegration test table

Posted: Mon Jul 23, 2012 10:33 am
by EViews Glenn
Can you be more precise about what options you used with coint? [edit] Since this post has been sitting here a while, while I wait for clarification, I want to point out that I've verified that the coefficient counts appear to be correct in the individual AIC, etc. in the VEC estimators.