Optimal lag length VECM

[Bjoern_1981]

Hey,

I hope my english is good enough to describe my problem.

For my paper I have to examine stochastical trends (multivariate cointegration of 7 different stock indices).

I dealt with the topics of unit root testing (ADF) and also a with the Johansen-Cointegration-test, both with Eviews 6.

A VECM-Modell was also estimated, but the optimal lag-length was not the same refering to the Johansen-Test (with lag length of 4 AIC gave me rank of cointegration of 1), but the VECM was optimal with a lag length of 3, which I found out by trying all possibilites and empirical tests.

Now my problem, because my professor told me, that there is an easier way to find the optimal lag. But I can't find how to do this. I read something about a "F-Test" but also "Testing the Autoregressive Order", for both I didn't find any explaining literature.

Thx for helping me...