I don't know how to get that particular error message from lmer, but
I can guess that it might occur when you are trying to estimate more
parameters than the data will support.
To overcome this, I would try the following:
First, have you tried experimenting with different models and / or
data sets to create very simple example examples, one that gives the
error you don't understand and another slighly different that seems to
work fine? That may be the easiest and most sensible thing to do. If
you get that far and still can't figure it out, I suggest you send that
simple, self-contained, reproducible example to this list. This is
something suggested in the posting guide!
(www.R-project.org/posting-guide.html). It can help you find the
problem and failing that can make it easier for someone else to
understand your problem. Without that, I'm just guessing, and I know
that some frequent contributors to this list are less likely to respond
to a question without something like that.
Beyond this, I could make other suggestions. For example, have you
tried method = "PQL" and "AGQ"? If both of these work, then
I would
think there might be something strange with "Laplace" at least for
this
example.
Also, I don't know what "family" you are using, but if it's
normal,
have you tried running the same model using lme in library(nlme)? (At
one time, I think you may have had to quit R and restart in changing
from lme4 to nlme or vice versa, I don't remember now, but I always do
that to avoid potential conflicts between the two packages.) If it's
NOT normal, I might still try the same model in lme ignoring the
nonnormality. If you get similar convergence problems or if
"intervals"
dies on you or give extremely wide confidence intervals, that suggests
you may be trying to estimate too many effects.
Is it convenient to make tables to find out how many combinations of
factors you have? If you slice things too thinly, the model can't be
estimated; you can get convergence problems from that.
If it were me, I might try to make a local copy of the code and walk
through it line by line using debug until I figured out the problem.
However, this might not be feasible if you're not familiar enough with
the algorithm to almost write your own code.
hope this helps,
spencer graves
Bill Shipley wrote:
> I am getting the following error message using the lmer function for mixed
> models with method="Laplace":
>
> "nlminb returned message false convergence (8) in:
LMEopt(x=mer,value=cv)"
>
> Could anyone explain what this means, and how I might overcome (or track
> down) the problem?
>
>
> Bill Shipley
>
>
>
>
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at 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