[R-sig-ME] pedigreemm with one observation per individual

[Aaron Rendahl]

I'm trying to fit a mixed model using pedigreemm, but unlike any of
the examples, I only have one observation per individual.
I'm getting the following error:

   Error in function (fr, FL, start, REML, verbose)  :
     Number of levels of a grouping factor for the random effects must
be less than the number of observations

which makes sense when the random effects are uncorrelated; as Spencer
Graves wrote when reporting this issue, "there are zero degrees of
freedom to distinguish 'group' from Residual."

However, with correlated random effects with a known correlation
pattern, as with pedigree data, it seems this is a reasonable thing to
do.  Is there a way to tell lme4 to allow this in this case?  Or is
this still not a reasonable thing to do?  Thanks for any comments or
suggestions you may have.

I'm running R 2.10.1, pedigreemm 0.2-4, and lme4 0.999375-32.

I've recreated the error using data from the milk dataset from
pedigreemm, as below.

> library(pedigreemm)
>
> # to simplify, just get one herd and remove unneeded columns
> milk <- within(milk, sdMilk <- milk / sd(milk))
> m<-subset(milk, herd==89)
> m<-m[,c("id","lact","dim","sdMilk")]
>
> # the data now has three measurements per cow
> head(m)
    id lact dim   sdMilk
1 6489    1 286 4.118756
2 6489    2 305 4.828023
3 6489    3 203 3.540520
4 6490    1 281 4.624768
5 6490    2 277 4.483228
6 6490    3 289 4.894880
>
> # and fitting a model works fine
> fm <- pedigreemm(sdMilk ~ lact + log(dim) + (1|id),
+   data = m, pedigree = list(id = pedCowsR))
>
> # but if I use only one measurement for each cow, as follows
> m1<-subset(m, lact==1)
> head(m1)
     id lact dim   sdMilk
1  6489    1 286 4.118756
4  6490    1 281 4.624768
7  6491    1 305 5.287302
8  6492    1 281 4.421290
12 6493    1 305 4.374557
16 6494    1 305 3.783129
>
> # it then fails
> fm1 <- pedigreemm(sdMilk ~ log(dim) + (1|id),
+   data = m1, pedigree = list(id = pedCowsR))
Error in function (fr, FL, start, REML, verbose)  :
  Number of levels of a grouping factor for the random effects
must be less than the number of observations

-- 
Aaron Rendahl, Ph.D.
Statistical Consulting Manager
School of Statistics, University of Minnesota

NEW OFFICE (as of June 2009):
48C McNeal Hall, St. Paul Campus
612-625-1062
www.stat.umn.edu/consulting