[R] An "unsubsettable object" in a mixed model

[Ben Bolker]

Sarah Jervis <sj414 <at> medschl.cam.ac.uk> writes:

> 
> I am having problems using the /lme /command to fit mixed models. I have 
> a data set similar to longitudinal data, except the hypothesised 
> correlation is between observations taken from different individuals in 
> the same family rather than from the same individual at different times. 
> As soon as I try to specify any correlation structure other than 
> independent, I get the error message "/Error in x$formula : object of 
> type 'closure' is not subsettable/". Have you any idea what /R /is 
> objecting to and how I might fix it, and if not, can you suggest h
> any other way I can get R to fit a linear mixed model? The data set is 
> VERY unbalanced with almost 2/3 of the "families" being single 
> individuals, so maybe /R/ doesn't like being told to find correlation in 
> univariate data subsets. ANY advice is welcome!

  It would help if you could provide a reproducible example, or
a test case.  You're also probably better off posting this to
the r-sig-mixed-model at r-project.org mailing list.  I don't know if
lme will choke when a correlation model is fitted to a data set where
some groups have only one individual, but you could (e.g.) very easily
test if this is the problem by (1) fitting the model with only groups
with n>1 and (2) adding one group with n=1 to the data set and seeing
if it chokes at that point.
  However, I can also imagine that you're mis-specifying the command
in some point, and from that point of view it would be best to have
some more detail about what you're trying to do.  See
http://tinyurl.com/reproducible-000 ...