R help - Apr 2007 - Example of mcmcsamp() failing with lmer() output

Hi,

I would appreciate help with the following model

<<1>>gunload <- read.table(hh('datasets/gunload.dat'),
header = T)
gunload$method <- factor(gunload$method, labels = c('new',
'old'))
gunload$physique <- factor(gunload$group, labels = c('slight',  
'average', 'heavy'))
gunload$team9 <- factor(rep(1:9, each = 2))
@

This is a split plot design with randomly selected teams for each  
physique, which was crossed with two methods.

Following the Oats example in chap. 1 of MEMSS, and adapting to lmer:

<<2>>require(lme4)
gl.lmer <- lmer(rounds ~ method * physique + (1 | physique/team),  
data = gunload)
mcmcsamp(gl.lmer)
@
lmer runs as expected, but mcmcsamp() gives:
Error: Leading minor of order 1 in downdated X'X is not positive  
definite
Error in t(.Call(mer_MCMCsamp, object, saveb, n, trans, verbose,  
deviance)) :
	error in evaluating the argument 'x' in selecting a method for  
function 't'

I tried:

<<3>>require(nlme)
gl.lme <- lme(rounds ~ method * physique, random = ~1 | physique/ 
team, data = gunload)
@
which gave all the expected output except that two p-values were NaN.

Which aspect of the data is making the simulation fail, and is there  
anything to be done about it?

_____________________________
Professor Michael Kubovy
University of Virginia
Department of Psychology
USPS:     P.O.Box 400400    Charlottesville, VA 22904-4400
Parcels:    Room 102        Gilmer Hall
         McCormick Road    Charlottesville, VA 22903
Office:    B011    +1-434-982-4729
Lab:        B019    +1-434-982-4751
Fax:        +1-434-982-4766
WWW:    http://www.people.virginia.edu/~mk9y/

On 4/27/07, Michael Kubovy <kubovy at virginia.edu>
wrote:
> Hi,
>
> I would appreciate help with the following model
>
> <<1>>> gunload <-
read.table(hh('datasets/gunload.dat'), header = T)
> gunload$method <- factor(gunload$method, labels = c('new',
'old'))
> gunload$physique <- factor(gunload$group, labels = c('slight',
> 'average', 'heavy'))
> gunload$team9 <- factor(rep(1:9, each = 2))
> @
>
> This is a split plot design with randomly selected teams for each
> physique, which was crossed with two methods.
>
> Following the Oats example in chap. 1 of MEMSS, and adapting to lmer:
>
> <<2>>> require(lme4)
> gl.lmer <- lmer(rounds ~ method * physique + (1 | physique/team),
> data = gunload)
> mcmcsamp(gl.lmer)
> @
> lmer runs as expected, but mcmcsamp() gives:
> Error: Leading minor of order 1 in downdated X'X is not positive
> definite
> Error in t(.Call(mer_MCMCsamp, object, saveb, n, trans, verbose,
> deviance)) :
>         error in evaluating the argument 'x' in selecting a method
for
> function 't'
I think you are overfitting the model - you have a fixed effect for
physique and a random effect for physique.  This is not the same as
the Oats example where there is a fixed effect for variety and a
random effect for variety within block, which designates the plot
within the block.

Try fitting the model with the random effects as  (1|physique:team)
and you will see that you get the same fitted values but you are able
to run mcmcsamp on this fit.

> I tried:
>
> <<3>>> require(nlme)
> gl.lme <- lme(rounds ~ method * physique, random = ~1 | physique/
> team, data = gunload)
> @
> which gave all the expected output except that two p-values were NaN.
>
> Which aspect of the data is making the simulation fail, and is there
> anything to be done about it?
>
> _____________________________
> Professor Michael Kubovy
> University of Virginia
> Department of Psychology
> USPS:     P.O.Box 400400    Charlottesville, VA 22904-4400
> Parcels:    Room 102        Gilmer Hall
>          McCormick Road    Charlottesville, VA 22903
> Office:    B011    +1-434-982-4729
> Lab:        B019    +1-434-982-4751
> Fax:        +1-434-982-4766
> WWW:    http://www.people.virginia.edu/~mk9y/
>
> ______________________________________________
> 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
> and provide commented, minimal, self-contained, reproducible code.
>