DW test

Discuss the jamovi platform, possible improvements, etc.

by mdelacre » Wed Jun 03, 2020 9:01 am

Hi,

When performin g a Linear Regression, for the Durbin-Watson test, the output returns same DW statistic and autorocorrelation, but unequal p, depending on whether the DV is centered or not.

jamovi centered2.jpg
jamovi centered2.jpg (46.29 KiB) Viewed 564 times


jamovi centered.jpg
With verreC=verres-VMEAN(verres)
jamovi centered.jpg (44.44 KiB) Viewed 564 times


With verresC=verres-VMEAN(verres)

Except for the p of the DW Statistic, all results are similar, whatever DV is centered or not. I don't know this test very well, but is there an explanation for that?

Sincerely,

Marie
mdelacre
 
Posts: 7
Joined: Fri Jul 12, 2019 9:39 am

by mcfanda@gmail.com » Fri Jun 05, 2020 4:16 pm

Hi
the reason for a different pvalue of the DW test is that jamovi (rightly) uses a bootstrap method to estimate it, so two different runs would produce two different results (try to run the same model with the same variables again and again and you'll see a change in the pvalue).
User avatar
mcfanda@gmail.com
 
Posts: 207
Joined: Thu Mar 23, 2017 9:24 pm

by mdelacre » Mon Jun 08, 2020 1:41 pm

okay, very cool, thanks!
mdelacre
 
Posts: 7
Joined: Fri Jul 12, 2019 9:39 am


Return to General