Skip to content

corARMA in lme (Arnaud Mosnier)

1 message · Highland Statistics Ltd

#
On 07/11/2012 13:00, r-sig-mixed-models-request at r-project.org wrote:
Arnaud...no you cannot conclude that. The random effect structure is 
already imposing a correlation on the data from the same tag, and also 
on data from the same yearTagged. Phi alone does not quantify your 
correlation. The ARMA is trying to add extra correlation. On top of 
that..the AR and ARMA correlation structures are quite specific.

If you think that there should be correlation, try different starting 
values for the phi.

Alain