CDtest (cross-sectional dependence test)

[EViews Esther]

This thread is about the CDtest add-in that lets you test for the cross-sectional dependence in panel data.

[Update] (07/01/2014) Added a new warning message, "*The distribution of Frees’ Q statistic requires T>3" for the Frees test
- Computation of the appropriate quantiles for the Frees' Q distribution would be available when the minimum number of periods T>3.

[gftiryaki]

Whenever I run any of the CDtests, I get the following error message:
PANCOV is not a valid view for_PRESID06 in "DO _PRESID06.PANCOV(OUT=_A01) UCOR OBS"
What could be the problem?

[EViews Gareth]

You need EV8.

[gftiryaki]

Ok, Thanks!

[tvonbrasch]

Hi,

Is there a bug in the CDtest addin?
in the following workfile, when I run the cdtest command on the equation object dols, 4 of the five test statistics show NA. Also, the test statistic Pearson LM Normal has the value 0.0000 yielding a probability of 1.0000. Can this be right? Is this a bug or is this right? Under what circumstances will thse test statistics typically yield NA?
Thomas

[EViews Esther]

There were computational issues because the given panel data is severely unbalanced (especially, *T<4*). Please reinstall your cdtest addin, and then you will see the following table with a warning message*.

The table shows the NA* value for the Frees test because computation of the appropriate quantiles of the Frees' Q distribution would be available for the case with T>3 only (where T is the minimum number of periods).

[tvonbrasch]

thank you for solving this issue!

[amarnath1991]

Dear Sir,

I have been trying to perform CD test in eviews 7 using the add in from your forum. But I encounter an error "PANCOV is not a valid view for_PRESID06 in "DO _PRESID06.PANCOV(OUT=_A01) UCOR OBS". There had been a previous post in the add ins forum where a person with Eviews 7 had got this problem resolved. Kindly let me know about how i can get over this problem.

[EViews Esther]

Can you upload your workfile so that I can replicate the error?

[EViews Glenn]

As you have found, the add-in does not work with EViews 7. It requires the panel covariance computation which is an EViews 8 feature.

[aawoad]

Hello
When I applied Residual Cross-Section Dependence Test on my unbalanced panel data (N=106, T=18), the finding for all types of tests are NA, could you please help?
Regards