[R-sig-ME] pedigreemm number of levels per grouping factor
Wilson, Alastair
A.Wilson at exeter.ac.uk
Thu Feb 6 16:50:45 CET 2014
Thanks for the suggestion Ben. Sadly it didn't provide a solution. The error message indicates the variance component I am after is unidentifiable, which would not be the case if I have correctly got the random ID effect linked to the pedigree.
Error in .sortCsparse(.Call(dtCMatrix_sparse_solve, a, b)) :
Dimensions of system to be solved are inconsistent
In addition: Warning message:
In checkZrank(reTrms$Zt, n = n, control, nonSmall = 1e+06) :
number of observations <= rank(Z); variance-covariance matrix will be unidentifiable
I'll maybe contact the package maintainer directly unless anyone else has thoughts/experience with pedigreemm.
Alastair
-----Original Message-----
From: r-sig-mixed-models-bounces at r-project.org [mailto:r-sig-mixed-models-bounces at r-project.org] On Behalf Of Ben Bolker
Sent: 05 February 2014 18:35
To: r-sig-mixed-models at r-project.org
Subject: Re: [R-sig-ME] pedigreemm number of levels per grouping factor
On 14-02-04 06:34 PM, Wilson, Alastair wrote:
> I’m trying to fit a pedigree-based mixed model (ie animal model) to
> estimate a genetic variance and a couple of fixed effects in a
> simulated data set. I’ve fitted the model using AsReml-R and
> MCMCglmm. I wanted to try using pedigreemm but get a error
>
> Error in checkNlevels(reTrms$flist, n = n, control) : number of levels
> of each grouping factor must be < number of observations
>
> I *think* this is telling me I don’t have multiple observations per
> factor level (here ID) which is true. However, I shouldn’t need
> repeat measures to estimate the random effect variance here (i.e.
> VA), so am assuming I have not got the code (below) right. Any
> pointers v gratefully received.
>
> Alastair
I'm not sure, but I guess that this is one of the cases where individual-level random effects *do* make sense in a LMM context. I don't know whether pedigreemm allows you to pass arguments through to lmer, but if it does you can use
control=lmerControl(check.nobs.vs.nlev="ignore") to suppress this check.
You might also be able to do this globally via
options(lmerControl=list(check.nobs.vs.nlev="ignore"))
(The maintainer of the package might want to consider suppressing this check by default, if this is a common issue ...)
Ben Bolker
>
> ############### library(pedigreemm) setwd("c:/prussia/")
>
> voles<-read.table("voles.txt", header=T)
> voles$ID<-as.factor(voles$ID) voles$sex<-as.factor(voles$sex)
>
> volesPED<-read.table("volesPED.txt",header=T)
> volesPED2<-pedigree(volesPED$SIRE,volesPED$DAM,volesPED$ID)
>
> model3<-pedigreemm(size~forage+sex+(1|ID),
> pedigree=list(ID=volesPED2), data=voles, REML=TRUE)
>
It would be nice if the example were reproducible ...
_______________________________________________
R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
More information about the R-sig-mixed-models
mailing list