s-news
[Top] [All Lists]

Re: Error in doubling

To: "paul hughes" <u7n96@hotmail.com>
Subject: Re: Error in doubling
From: Douglas Bates <bates@stat.wisc.edu>
Date: 28 Feb 2001 10:20:15 -0600
Cc: s-news@lists.biostat.wustl.edu
In-reply-to: "paul hughes"'s message of "Wed, 28 Feb 2001 14:36:14 -0000"
References: <F226S97Q98fzimZSAEf0001a392@hotmail.com>
User-agent: Gnus/5.0803 (Gnus v5.8.3) Emacs/20.7
"paul hughes" <u7n96@hotmail.com> writes:

> Once again I have a simple problem that I need help on. When I try to fit 
> the mixed effects model to a data set:
> 
> armd.fit1<-lme(fixed=y~ti+tims+trt,cluster=~ID,
> data=real.data,na.action = na.omit)
> 
> I get this error:
> 
> armd.fit1 <- lme(fixed = y ~ ti + tims + trt, cluster =  ~ ID, data =
>       real.data, na.action = na.omit)
> Error in double(q * q * M): Cannot allocate 29478400 bytes: 
> options("object.size") is 5000000: see options help file
> Dumped

The "double" referred to in the error message is the function that
allocates a vector with storage mode double.  I think the problem
comes from the fact that you did not specify a random-effects model and
the default is the same as the fixed-effects model.  That will make
"q", the number of columns in the random-effects model matrices,
large.  The M is the number of distinct values of the "ID" factor.

You probably want to specify fewer random-effects in the model.

<Prev in Thread] Current Thread [Next in Thread>