The tests you rune below are not identical.
For the regression, the test is an F-test based on the residuals of the
one equation regression. It is not a test of granger non-causality.
For the vargranger test, the test is a Chi-squared test based on the
residuals of both equations of the two equation var system.
In the first case, you should have run a var and then tested the
coefficients of variable2 in the variable1 equation.
Even then you may not get the same results if Stata produces an F-test
instead of a Chi-squared test in this case. I have not tried it.
Eric de Souza
College of Europe
Brugge (Bruges), Belgium
http://www.coleurope.eu
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Martin
Baumann
Sent: 09 July 2009 11:34
To: [email protected]
Subject: st: Granger Causality Test after regress different from
vargranger
Dear all
When testing for Granger Causality I get slightly different p-values
when using a regress and test command (approach1) as opposed to using a
var and vargranger command (approach2). Are these two approaches
identical as the Time Series Manuals suggests or is there a theoretical
difference between the two? If yes, which one would be preferrable in a
single equation framework?
Thanks a lot for any help. Best regards,
Martin
Approach 1: Classical Granger Causality Test
-------------------------------------------------------------------
reg d.Variable1 D.L(1/10).Variable2 D.L(1/10).Variable1
test D.L.Variable2 D.L2.Variable2 .... D.L10.Variable2
Approach 2: Granger Test with var-command
---------------------------------------------------------------------
var d.Variable1 d.Variable2, lags(1/10)
vargranger
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
