R help - Dec 2007 - difficulties getting coef() to work in some lmer() calls

I'm working with Andrew Gelman on a book project and we're having some
difficulties getting coef() to work in some lmer() calls.

Some versions of the model work and some do not.  For example, this works
(in that we can run the model and do coef() from the output):

R2 <- lmer(y2 ~ factor(z.inc) + z.st.inc.full + z.st.rel.full + (1 + factor(
z.inc) | st.num), family=binomial(link="logit"))

But this does not (the model runs but coef() doesn't work):

R3 <- lmer(y2 ~ factor(z.inc) + z.st.inc.full + z.st.rel.full + (1 + z.inc |
st.num ), family=binomial(link="logit"))

We get the following error:
Error in coef(R3) : unable to align random and fixed effects

The data are at
http://www.stat.columbia.edu/~gelman/temp/lmerexample.txt<http://www.stat.columbia.edu/%7Egelman/temp/lmerexample.txt>
The file to read the data are at:
http://www.stat.columbia.edu/~gelman/temp/lmerexample.R<http://www.stat.columbia.edu/%7Egelman/temp/lmerexample.txt>

Any help would be appreciated.

David K. Park

-- 
David K. Park
Visiting Researcher (Fall 2007 / Spring 2008)
Applied Statistics Center
Columbia University
1016 Social Work Bldg
(Amsterdam Ave. at 122 St.)
New York, NY 10027

Assistant Professor
Department of Political Science
George Washington University
Monroe Hall
2115 G Street, NW
Washington, D.C. 20052
http://home.gwu.edu/~dkp/ <http://home.gwu.edu/%7Edkp/>

Direct: (202) 994-2331
Dept: (202) 994-6290
Fax: (202) 994-7743

"The great tragedy of Science - the slaying of a beautiful hypothesis
by an ugly fact."
-- Thomas H. Huxley

	[[alternative HTML version deleted]]

On Dec 3, 2007 1:45 PM, David Park <dkp at gwu.edu>
wrote:
> I'm working with Andrew Gelman on a book project and we're having
some
> difficulties getting coef() to work in some lmer() calls.
>
> Some versions of the model work and some do not.  For example, this works
> (in that we can run the model and do coef() from the output):
>
> R2 <- lmer(y2 ~ factor(z.inc) + z.st.inc.full + z.st.rel.full + (1 +
factor(
> z.inc) | st.num), family=binomial(link="logit"))
>
> But this does not (the model runs but coef() doesn't work):
>
> R3 <- lmer(y2 ~ factor(z.inc) + z.st.inc.full + z.st.rel.full + (1 +
z.inc |
> st.num ), family=binomial(link="logit"))
>
> We get the following error:
> Error in coef(R3) : unable to align random and fixed effects
Which indicates that the coefficients being estimated for each level
of st.num do not correspond to any of the fixed-effects coefficients,
hence the coef method, in the sense that it is used in nlme and lme4,
cannot be applied.

You will need to change the fixed-effects specification so that it
includes z.inc.  This may mean modifying the contrasts from
factor(z.inc).  I would suggest switching to polynomial contrasts then
replacing the linear term by z.inc
> The data are at
>
http://www.stat.columbia.edu/~gelman/temp/lmerexample.txt<http://www.stat.columbia.edu/%7Egelman/temp/lmerexample.txt>
> The file to read the data are at:
>
http://www.stat.columbia.edu/~gelman/temp/lmerexample.R<http://www.stat.columbia.edu/%7Egelman/temp/lmerexample.txt>
>
> Any help would be appreciated.
>
> David K. Park
>
> --
> David K. Park
> Visiting Researcher (Fall 2007 / Spring 2008)
> Applied Statistics Center
> Columbia University
> 1016 Social Work Bldg
> (Amsterdam Ave. at 122 St.)
> New York, NY 10027
>
> Assistant Professor
> Department of Political Science
> George Washington University
> Monroe Hall
> 2115 G Street, NW
> Washington, D.C. 20052
> http://home.gwu.edu/~dkp/ <http://home.gwu.edu/%7Edkp/>
>
> Direct: (202) 994-2331
> Dept: (202) 994-6290
> Fax: (202) 994-7743
>
> "The great tragedy of Science - the slaying of a beautiful hypothesis
> by an ugly fact."
> -- Thomas H. Huxley
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>