[R] What computer power for GAMM models

[Bert Gunter]

Sylvie:

I do not know whether your system lacks sufficient power, but its user
-- you --  may.

The models you are trying to fit are complex and nonlinear, and
possibly depending on your data -- do you have a lot of missingness?
How large are your data sets? Can they support your models -- you may
well be overfitting. Getting a faster system will not affect this --
you'll just get your error messages faster.

... and I would have thought the missing values error is self-explanatory.

Sounds to me like you should spend some time with a local statistical expert.

Cheers,
Bert

On Fri, Oct 25, 2013 at 6:19 AM, Sylvie Martin <smartin_sepia at orange.fr> wrote:
>
>
> Hello,
>
>  I use GAMM function (MGCV package) in R software to study relationship between pollen and pollinosis. My models include autoregressive terms. Here is an exemple:
>
> CupAR7_2<-gamm(nb_rca_2~s(Trend,bs="ps",k=49) +Semaine
>             +s(TX03,bs="ps",k=8) +s(UN03,bs="ps",k=4)  ......+pollen01
> ,correlation=corARMA(p=7,q=0,form=~Trend|Annees),
> niter=40,control=lmeControl(maxIter=100000,msMaxIter=100000)
>         ,family=quasipoisson, data=donnees, na=na.omit)
>
> Each model need 1/2 hour to 1 hour to get a result and I frequently get the following error messages:
>
> Erreur dans lme.formula(fixed = fixed, random = random, data = data, correlation = correlation,  :
>   nlminb problem, convergence error code = 1
>   message = singular convergence (7)
>
> Erreur dans na.fail.default(list(Xr.4 = c(0.00374377951791214, 0.00438373736920182,  :
>   missing values in object
>
>
>  Does my computer lack power,(if so,what is required?) or is R  limited in speed of execution or is my syntax not ok. Here are the features of my computer:
>
> processor: 3.10 GHz  i5-2400 CPU
> RAM: 4 Go (3.87 Go)
> OS: 64 bits
> Windows 7 professionnel (Pack 1)
>
> Another question: how can we add a smoothing parameter in the model GAMM to help convergence ("sp =" is not accepted)?
>
> Thanks for your help
> Best regards
>
> Sylvie Martin
> SEPIA-Santé
>
> Bureau d'études en épidémiologie,
> biostatistiques, environnement
>
> 31 rue de Pontivy
>
> 56 150 BAUD
>
> FRANCE
> tél : 02 97 28 80 38
> fax: 02 97 28 81 10
>         [[alternative HTML version deleted]]
>
>
> ______________________________________________
> 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.
>



-- 

Bert Gunter
Genentech Nonclinical Biostatistics

(650) 467-7374