Dear S+ users:
I'am using for the first time S-Plus 4.0 under Win95, and I'am working with
Generalized
Additive Models.
After having estimated the model using the function gam, I can get an
anova-like table showing me, as is reported in the on-line manual,
the non-linear effect of each predictor included in the model.
I have to explain the effect of about 30 predictors over my response
variable, and when I increase the number of predictor entered I note
that some of them entered before are not significant anymore
(i.e. they have associated an high F or Chisq value).
My first question is:
Is there any criterion to set the order of the predictor to be
included in the model?
Furthermore, I used the following expression of the function predict.gam
in order to predict my new values of the dependent variable, supplying
a "newdata" object.
mib1.gam.pred.new<-
predict.gam(mib1.gam,mibp1gam,type="response")
I get the following error message:
Error in .Fortran("lowese",: extrapolation not allowed with blending
Dumped
Warning messages:
1: eval MbPjt? -0.085908
in: .Fortran("lowese", ...
2: lowerlimit eval MbPjt? -0.07646
in: .Fortran("lowese", ...
Could anyone explain me what does it mean?
Thanks in advance.
Claudio
,,,
(o o)
--oOO--(_)--OOo-------------------------------------------------------
Claudio Conversano
Ph.D. Student
Dipartimento di Matematica e Statistica
Via Cintia Monte S. Angelo
I-80126 NAPOLI
ITALY
-----------------------------------------------------------------------
-----------------------------------------------------------------------
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
|