The degrees of freedom of the asymptotic distribution of the
Ljung-Box test statistic is the number of auto correlations included
minus the number of parameters estimates. You will find an
explanation in any good time series test (e.g. Brockwell and Davis
(2006), Shumway and Stoffer (2006 2011(?)), Tsay (2010) ). Remember
this result is only asymptotic and therefore is only approximate.
Some software packages therefore do not include the adjustment to the
degrees of freedom. There are no universally accepted rules as to
the number of lags that should be included in the autocorrelation
function.
Best Regards
John
2011/5/23 Sam Sam <dear.sam(a)livemail.tw>:
Dear all:
I am confused about that degree of freedom is the number of lag or the nuber
of lag minus the number of parameter estimated
(When using Ljung-Box Q stat. to test if the residuals of ARIMA is
autocorrelated.)
It seems that degree of freedom of the nuber of lag minus the number of
parameter estimated used in many study.
Thanks a lot
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
--
John C Frain
Economics Department
Trinity College Dublin
Dublin 2
Ireland
www.tcd.ie/Economics/staff/frainj/home.html
mailto:frainj@tcd.ie
mailto:frainj@gmail.com