Tomas,
I suggest that you use
try()
Andrew
On Wed, Apr 30, 2008 at 05:30:05PM +0200, Tomas Goicoa wrote:
> Dear R users,
>
> I want to conduct a small simulation study and I have to use the lme
> function in a loop to save the restricted log likelihood.
> However, for one simulated data set the lme function gives this error
>
>
> Error en lme.formula(yboot ~ X[, -1], data = data.fr, random = Z.block) :
> nlminb problem, convergence error code = 1
> message = singular convergence (7)
>
> and then, the simulation stops. I would like to skip this
> simulated data, and then continue with the loop, but I cannot find
> the way of doing this. I am using R2.6.2.
>
> Can anybody tell me if I can write a condition to avoid the
"problematic" data?
>
> I Have also obtained similar errors
>
> Error en lme.formula(yboot ~ X[, -1], data = data.fr, random = Z.block) :
> nlminb problem, convergence error code = 1
> message = false convergence (8)
>
>
>
>
> My data can be reproduced as follows
>
>
> library(splines)
> library(nlme)
>
>
> x<-c(rep(1993:2007,9))
> barrio<-factor(c(rep(1:9,rep(15,9))))
>
> xl<-min(x)-0.00001
> xr<-max(x)+0.00001
> ndx<-4
> bdeg<-3
> pord<-2
> dx <- (xr-xl)/ndx
> knots <- seq(xl-bdeg*dx, xr+bdeg*dx, by=dx)
>
> B<-spline.des(knots,x,bdeg+1,0*x)$design
>
> m=ncol(B)
> D=diff(diag(m),differences=pord)
> P=t(D)%*%D
> P.svd=svd(t(D)%*%D)
> U=(P.svd$u)[,1:(m-pord)]
> d=(P.svd$d)[1:(m-pord)]
> Delta=diag(1/sqrt(d))
> Z=B%*%U%*%Delta
> X=NULL
> for(i in 0:(pord-1)){X=cbind(X,x^i)}
>
> ### This X and Z matrix are the matrices in the mixed model
> representation of a spline
>
>
> n<-nrow(X)
> Id<- factor(rep(1,n))
> Z.block<-list(list(Id=pdIdent(~Z-1)),list(barrio=pdIdent(~1)))
> Z.block<-unlist(Z.block,recursive=FALSE)
>
> fijos2<-c(-326.9178203,0.1645375 )
> ef.spline<-c(-0.2214524, -0.1649163, -0.1649163, 0.8878776, -0.2214524)
>
> set.seed(48)
> ef.error<-rnorm(135,0,sqrt(0.03073974))
>
> where
>
> ef.spline are standardized random effects, and errores are the residuals.
>
>
>
>
> My loop looks like this
>
>
> B<-1000
> y.boot<-array(0,c(nrow(Z),B))
> log.comp<-array(0,B)
>
>
> while(i<B){
> i<-i+1
> spline<-sample(ef.spline,size=ncol(Z),replace=T)
> errores<-sample(ef.error,size=nrow(Z),replace=T)
> y.boot[,i]<-X%*%fijos2+Z%*%spline+errores
> yboot<-y.boot[,i]
> data.fr <- groupedData( yboot ~ X[,-1] |Id ,data =
> data.frame(yboot,X,Z,barrio))
> log.comp[i]<-lme(yboot~X[,-1],data=data.fr,random=Z.block)$logLik
> }
>
>
>
> The idea is to add a condition in the loop such that
>
> if(condition......) {i<-i-1}
>
> and then substitute the "problematic" data by a new one,
>
> Many thanks,
>
> Tomas Goicoa
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help@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.
--
Andrew Robinson
Department of Mathematics and Statistics Tel: +61-3-8344-6410
University of Melbourne, VIC 3010 Australia Fax: +61-3-8344-4599
http://www.ms.unimelb.edu.au/~andrewpr
http://blogs.mbs.edu/fishing-in-the-bay/
