[S] modifying df for glm statistics

Carlos Carroll (carlos@pcweb.net)
Mon, 06 Jul 1998 14:36:44 -0700


To gauge the effects of spatial autocorrelation in a data set on the
significance values of regression coefficients, I am using a Fortran
program (based on the CRH method [ Cliford et al. 1989 ]) that computes
a corrected sample size based on the spatial covariance of the data
(reduced sample size in cases of positive autocorrelation).
I would like to use this number as adjusted degrees of freedom
to correct the significance of coefficients in a glm object, but I
realized that the Chisquared statistic given by anova.glm and the
t-value given by summary.glm are already based on a deviance value that
depends on the uncorrected
degrees of freedom, so it's not as simple as calling pchisq() with a
new df. Can anyone suggest a way to modify the glm.object (perhaps the
object$R) to account for reduced df and produce revised t or Chisq
values?
Apologies if this is a basic question.
Thanks in advance for any assistance.
-- 
Carlos Carroll, M.S.
Conservation Biology Institute and
Dept. of Forest Science, Oregon State University
Mailing address: PO Box 104
Orleans, CA 95556
e-mail: carlos@pcweb.net
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu.  To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message:  unsubscribe s-news