## Cointegration Test gives wrong P-Value

For notifying us of what you believe are bugs or errors in EViews.

Moderators: EViews Gareth, EViews Moderator

grabodan
Posts: 7
Joined: Thu Jun 11, 2015 6:36 am

### Cointegration Test gives wrong P-Value

Dear EViews-Team
When I perform a Johansen Cointegration Test for a VAR and the Trace od ME Test Statistic are much larger than the critical value, the p-value should be close to zero. Instead, I shows the Prob: 1.0000.
The programm code I used, just as an example, was:

Code: Select all

`wfcreate(wf=Koint) u 10000smpl @allseries x=0series y=0series z=0series trend = 0smpl @first+1 @lastseries u=@nrndseries v=@nrndseries w=@nrndtrend = 1+trend(-1)+@nrndx= trend + uy= y(-1) + 0.5*(x(-1) - y(-1))  + vz= 10 + z(-1) + wvarest 1 1 x y z`

I am using EViews 10, but I think I have encountered the same problem already years ago with Version 8 or so.

Edit: The issue disappears if, e.g., I reduce the number of obs to 1000 such that the p-value is 0.0001 or such. Thus, I only appears for p-values extreme close to zero, displaying 1 instead.

Best regards,
Daniel
Attachments
koint.wf1

EViews Glenn
EViews Developer
Posts: 2600
Joined: Wed Oct 15, 2008 9:17 am

### Re: Cointegration Test gives wrong P-Value

Thanks. We will take a look.

EViews Glenn
EViews Developer
Posts: 2600
Joined: Wed Oct 15, 2008 9:17 am

### Re: Cointegration Test gives wrong P-Value

We've identified the issue.

In the MacKinnon algorithm for the response surface, we have found that there are potential issues with the polynomial approximation to the response surface for extreme values. In this case, the statistic value is so out of bounds that the existing algorithm approximation is quite poor. We have implemented a fix for the tail conditions which provides additional robustness to these large values and are in the process of testing the changes. A fix will appear soon (most likely in the next patch for EViews 10).

grabodan
Posts: 7
Joined: Thu Jun 11, 2015 6:36 am

### Re: Cointegration Test gives wrong P-Value

Thanks for the response,
Daniel