@JOHMLE Question

[timduy]

Quick question because I see this presented in different ways. Suppose the output from the @JOHMLE procedure is:

Code: Select all

Unrestricted eigenvalues and -T log(1-lambda)
   Rank     EigVal   Lambda-max  Trace  Trace-95%   LogL
         0                                        1198.3102
         1    0.1068    16.4915 17.7699   15.4100 1206.5560
         2    0.0087     1.2784  1.2784    3.8400 1207.1952

I assume that "Rank 1" means a test of the null that the rank is 0 versus the alternative that the rank is at least 1 and "Rank 2" is the null that the rank is less than or equal to one against the alternative that the rank is 2. So then "Rank" here is essentially the alternative hypothesis, so that in this case there is one cointegrating vector.

[TomDoan]

That's correct.

[abi]

Hi Tom,

Unlike trace statistic Critical value for Lambda-max are not reported in the result, is there any way by which i can get them?

Thanks,

[TomDoan]

You use the trace statistic to decide the rank. There's no real point in critical values for lambda-max. (CATS doesn't even include the lambda-max statistic, which is just a rescale of the eigenvalue).

[abi]

Thank you Tom for quick reply,

Actually i found that EViews reports both trace and lambda-max statistics, so my question came from there.

[TomDoan]

Johansen and Juselius don't think it's even worth reporting. I trust them on that.