I have two time series, for temperature and precipitation, and a fitted
model that is a function of T, P and T*P. I would like to deseasonalize
the independent variables from observed data so that they have zero mean,
obtain an ar multivariate model, then add back the seasonality and
simulate a time series of the dependent variable. The problem is that,
since T and P are correlated its time series does not have zer mean.
What to do? The problem appears to be the way I am using ar(), but I
can't find descriptions of the multivariate ar to help.
Thanks in advance.
--s
-------------------------------------------------------------------------
Steven C. Wofsy
Abbott Lawrence Rotch Professor of Atmospheric and Environmental Chemistry
Harvard University
Division of Engineering and Applied Science/Department of Earth and
Planetary Science
29 Oxford St., Cambridge, MA 02138
Tel. 617-495-4566; FAX 617-495-4551; http://www-as.harvard.edu
|