R help - Jun 2010 - varIdent error using gam function in mgcv

Hello,

As I am relatively new to the R environment this question may be either
a) Really simple to answer
b) Or I am overlooking something relatively simple.

I am trying to add a VarIdent structure to my gam model which is fitting
smoothing functions to the time variables year and month for a particular
species. When I try to add the varIdent weights to variable Month I get this
error returned. 

Error in model.frame.default(formula = Chel ~ 1 + M + YM + M + YM, weights Vf1, 
:
  variable lengths differ (found for '(weights)')

Here is the simple code I am using.
library(mgcv); library(lattice)
library(nlme)
Vf1=varIdent(form=~M)
Chel2=gam(Chel~s(M, fx=F, bs="tp")+ s(YM, fx=F,
bs="tp")+s(M,YM, fx=F,
bs="tp"), family=poisson(link="log"), weights=Vf1) # second
GAM  with
interaction

M=Month
YM=Year

Regards,
Niall
-- 
View this message in context:
http://r.789695.n4.nabble.com/varIdent-error-using-gam-function-in-mgcv-tp2260556p2260556.html
Sent from the R help mailing list archive at Nabble.com.

How exactly do you want to define your weights? varIdent returns a
varFunc object, not a vector with the same length as the data
variables, as required by the gam function.

Cheers
Joris

On Fri, Jun 18, 2010 at 7:35 PM, niallllll <n.mcginty1 at nuigalway.ie>
wrote:
>
> Hello,
>
> As I am relatively new to the R environment this question may be either
> a) Really simple to answer
> b) Or I am overlooking something relatively simple.
>
> I am trying to add a VarIdent structure to my gam model which is fitting
> smoothing functions to the time variables year and month for a particular
> species. When I try to add the varIdent weights to variable Month I get
this
> error returned.
>
> Error in model.frame.default(formula = Chel ~ 1 + M + YM + M + YM, weights
> Vf1, ?:
> ?variable lengths differ (found for '(weights)')
>
> Here is the simple code I am using.
> library(mgcv); library(lattice)
> library(nlme)
> Vf1=varIdent(form=~M)
> Chel2=gam(Chel~s(M, fx=F, bs="tp")+ s(YM, fx=F,
bs="tp")+s(M,YM, fx=F,
> bs="tp"), family=poisson(link="log"), weights=Vf1) #
second GAM ?with
> interaction
>
> M=Month
> YM=Year
>
> Regards,
> Niall
> --
> View this message in context:
http://r.789695.n4.nabble.com/varIdent-error-using-gam-function-in-mgcv-tp2260556p2260556.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>


-- 
Joris Meys
Statistical consultant

Ghent University
Faculty of Bioscience Engineering
Department of Applied mathematics, biometrics and process control

tel : +32 9 264 59 87
Joris.Meys at Ugent.be
-------------------------------
Disclaimer : http://helpdesk.ugent.be/e-maildisclaimer.php

On Fri, 2010-06-18 at 10:35 -0700, niallllll wrote:
> Hello,
> 
> As I am relatively new to the R environment this question may be either
> a) Really simple to answer
Use gamm() not gam()
> b) Or I am overlooking something relatively simple.
Use gamm() not gam() ;-)
> I am trying to add a VarIdent structure to my gam model which is fitting
> smoothing functions to the time variables year and month for a particular
> species. When I try to add the varIdent weights to variable Month I get
this
> error returned. 
> 
> Error in model.frame.default(formula = Chel ~ 1 + M + YM + M + YM, weights
> Vf1,  :
>   variable lengths differ (found for '(weights)')
In gam(), weights is an argument where you need to specify the numeric
prior weights for the observations. If you have n observations then
length(weights) == n.

varIdent is a variance function from the nlme package. It specifies a
variance function, the parameters of which will be estimated as part of
the model fitting, but only within a call to lme (and/or nlme I
presume).

To use it, you'll need to use the gamm() function which will eventually
use lme (via glmmPQL in package MASS). Whether this is something you
want to do is another matter.

I'm not even sure that 'weights' in glm/gam should be used in the
way
you are trying to; I recall some discussion on this list that weights
are for sampling weights, but I may be mis-remembering.

Regardless, you can't use varIdent like that in a gam() model.

HTH

G
> Here is the simple code I am using.
> library(mgcv); library(lattice)
> library(nlme)
> Vf1=varIdent(form=~M)
> Chel2=gam(Chel~s(M, fx=F, bs="tp")+ s(YM, fx=F,
bs="tp")+s(M,YM, fx=F,
> bs="tp"), family=poisson(link="log"), weights=Vf1) #
second GAM  with
> interaction
> 
> M=Month
> YM=Year
> 
> Regards,
> Niall
-- 
%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%
 Dr. Gavin Simpson             [t] +44 (0)20 7679 0522
 ECRC, UCL Geography,          [f] +44 (0)20 7679 0565
 Pearson Building,             [e] gavin.simpsonATNOSPAMucl.ac.uk
 Gower Street, London          [w] http://www.ucl.ac.uk/~ucfagls/
 UK. WC1E 6BT.                 [w] http://www.freshwaters.org.uk
%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%