Re: [R] 'nlme' crashes R (was: Using corStruct in nlme)

From: Spencer Graves <spencer.graves_at_pdf.com>
Date: Fri 21 Jul 2006 - 03:43:27 EST

          Thanks for providing such a self-contained example by which 'nlme' crashes R. Could you please also give us 'sessionInfo()'? I don't have time to test it myself now, but perhaps if you identify your platform, you might interest someone else in checking it.

	  I'm sorry I couldn't be more helpful.
	  Spencer Graves

grieve@u.washington.edu wrote:
> I am having trouble fitting correlation structures within nlme. I would like to
> fit corCAR1, corGaus and corExp correlation structures to my data. I either
> get the error "step halving reduced below minimum in pnls step" or
> alternatively R crashes.
>
> My dataset is similar to the CO2 example in the nlme package. The one major
> difference is that in my case the 'conc' steps are not the same for each 'Plant'.
I have replicated the problem using the CO2 data in nlme (based off of the Ch08.R
script).
>
> This works (when 'conc' is the same for each 'Plant':
>
> (fm1CO2.lis <- nlsList(SSasympOff, CO2))
> (fm1CO2.nlme <- nlme(fm1CO2.lis, control = list(tolerance = 1e-2)))
> (fm2CO2.nlme <- update(fm1CO2.nlme, random = Asym + lrc ~ 1))
> CO2.nlme.var <- update(fm2CO2.nlme,
> fixed = list(Asym ~ Type * Treatment, lrc + c0 ~ 1),
> start = c(32.412, 0, 0, 0, -4.5603, 49.344),
> weights=varConstPower(fixed=list(const=0.1, power=1)), verbose=T)
>
> CO2.nlme.CAR<-update(CO2.nlme.var, corr=corCAR1())
>
> CO2.nlme.gauss<-update(CO2.nlme.var,
> correlation=corGaus(form=~as.numeric(conc)|Plant,nugget=F), data=CO2)
>
> CO2.nlme.exp<-update(CO2.nlme.var,
> correlation=corExp(form=~as.numeric(conc)|Plant,nugget=F), data=CO2)
>
> But, if i change each of the 'conc' numbers slightly so that they are no longer
identical between subjects i can only get the corCAR1 correlation to work while R
crashes for both corExp and corGaus:
>
> for(i in 1:length(CO2$conc)){
> CO2$conc[i]<-(CO2$conc[i]+rnorm(1))
> }
>
> (fm1CO2.lis <- nlsList(SSasympOff, CO2))
> (fm1CO2.nlme <- nlme(fm1CO2.lis, control = list(tolerance = 1e-2)))
> (fm2CO2.nlme <- update(fm1CO2.nlme, random = Asym + lrc ~ 1))
> CO2.nlme.var <- update(fm2CO2.nlme,
> fixed = list(Asym ~ Type * Treatment, lrc + c0 ~ 1),
> start = c(32.412, 0, 0, 0, -4.5603, 49.344),
> weights=varConstPower(fixed=list(const=0.1, power=1)), verbose=T)
>
> CO2.nlme.CAR<-update(CO2.nlme.var, corr=corCAR1())
>
> CO2.nlme.gauss<-update(CO2.nlme.var,
> correlation=corGaus(form=~as.numeric(conc)|Plant,nugget=F), data=CO2)
>
> CO2.nlme.exp<-update(CO2.nlme.var,
> correlation=corExp(form=~as.numeric(conc)|Plant,nugget=F), data=CO2)
>
> I have read Pinheiro & Bates (2000) and i think that it should be possible to fit these correlation structures to my data, but maybe i am mistaken.
>
> I am running R 2.3.1 and have recently updated all packages.
>
> Thanks,
> Katie Grieve
>
> Quantitative Ecology & Resource Management
> University of Washington
>
> ______________________________________________
> R-help@stat.math.ethz.ch 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.



R-help@stat.math.ethz.ch 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. Received on Fri Jul 21 03:58:05 2006

Archive maintained by Robert King, hosted by the discipline of statistics at the University of Newcastle, Australia.
Archive generated by hypermail 2.1.8, at Fri 21 Jul 2006 - 06:19:25 EST.

Mailing list information is available at https://stat.ethz.ch/mailman/listinfo/r-help. Please read the posting guide before posting to the list.