[R-sig-ME] Error message with lme

[FMH]

Thank you



----- Original Message ----
From: Douglas Bates <bates at stat.wisc.edu>
To: FMH <kagba2006 at yahoo.com>
Cc: r-sig-mixed-models at r-project.org
Sent: Wednesday, September 9, 2009 2:08:10 PM
Subject: Re: [R-sig-ME] Error message with lme

On Wed, Sep 9, 2009 at 7:16 AM, FMH <kagba2006 at yahoo.com> wrote:
> Dear All,
>
> I tried to run an R scipt on a set of data with lmList function and it worked fine, but as i tried to run with lme function, the program suddently give a message error. Both script and message error are  shown below
>
> #################################################################################
>> dp <- rep(rev(1:11), 261)
>> sub1 <- data.frame(sub1, dp, or)
>> tmp <- groupedData(Temp ~ dp  | or, data = sub1, FUN = mean, order.groups = TRUE, labels = list(x = "Depth", y =
>     "Temperature"),  units = list(y = "(0C)"))
>> lm.lis1 <- lmList(tmp)
>> lm.lme1 <- lme(lm.lis1)
>
> Error in lme.formula(fixed = Temp ~ dp, data = tmp, random = list(or = c(2.45530618001172,  :
>   nlminb problem, convergence error code = 1
>   message = iteration limit reached without convergence (9)
> #################################################################################
>
> Could someone advice me the way to tackle on this problem?

The first thing to do is to request verbose output to see what the
optimization is doing.  Use control = list(msVerbose = 1) in the call
to lme.

You may meet with more success using lmer from the lme4 package.  The
lmer function is a later design that takes advantage of further
research into linear mixed-effects models.  In particular, if the
problem you are encountering is due to a singular variance-covariance
matrix it will be handled much more effectively in lmer than in lme.