warning message for lmer model with poisson family

Looks like you have some independent
variables in your model that correlate
perfectly so your design matrix is
not of full rank, probably because
your independent variable data is not
balanced.

Does a simpler fit such as
  model<-lmer(count~tem+(1|id),family=poisson)
give you a result?

You might need to build up your model from such
a simple one, adding one new term at a time
until you find the one that causes the issue.

You probably do not have enough balanced
data to allow fitting such a complex
model.

HTH

Steve McKinney

-----Original Message-----
From: r-help-bounces at r-project.org on behalf of arams
Sent: Wed 6/4/2008 5:18 PM
To: r-help at r-project.org
Subject: [R]  warning message for lmer model with poisson family

Hello,

I'm trying to run an lmer model with family poisson but receive the
following
warning message: 

model<-lmer(count~tem+dat+alt+year+tem:dat+tem:alt+tem:year+dat:alt+dat:year+alt:year+(1|id),family=poisson)
Error in objective(.par, ...) : 
  Leading minor of order 2 in downdated X'X is not positive definite

I'm using the newest version of R and recently installed the following
libraries: (lme4, matrix, lattice).
I don't understand what the warning message means or know how to fix it,
please help!
My response variable is a count variable (n=551); 212 with value = 0, and
the rest range between 1-3.

warning message for lmer model with poisson family

Thread (2 messages)