influence values in the context of GLM

[impromptu@nifty.com]

S-Plus users
E-mail: s-news@wubios.wustl.edu

Hi!

   I am constructing a Fortran code to implement
log-linear models.
   I am just wondering what is the definition of influence values in
the context of log-linear models. The influence values(infl) are 
obtained by the S-Plus object below:

  data2 <- data.frame(x = xx, y = yy, www = wta)
  fit.glm <- glm(y ~ x, data = data2, family = poisson,
  weights = www, qr = T, x = T, y = T, model = T)
  infl <- lm.influence(fit.glm)

where "wta" is the weight of each data(xx,yy).
   These influence values seem to be different from
those derived form the diagonal elements of Hat matrix defined on 
page 397 in "Generalized Linear Model(by P.McCullagh and J.A.
Nelder)".

   *****    Kunio Takezawa, Ph.D. (takezawa@affrc.go.jp)    *****
*****     <http://cse.inada.affrc.go.jp/takezawa/patent-e.html>    *****