[R] need help on nlme()

[Prof Brian Ripley]

On Mon, 20 Feb 2006, ronggui wrote:

> I think nlme is not for logistic mixed effect model.
> you should use glmmPQL in MASS or lmer in Matrix

There are two senses of 'logistic mixed effect model'.  One is for a 
continuous response as given by SSlogis, and nlme is appropriate.  The 
other is a glm with a binomial (or binary) response, and nlme is 
inappropriate.

The issue here is that when the fixed includes an rhs, there are 
parameters for intercept and slope for each level of SKILLS, although they 
will be coded using contrasts.  So 2 starting values are not enough.

>
> 2006/2/20, Mingyu Feng <mingyufeng at gmail.com>:
>> Hello there,
>>
>> I am using nlme() to fit a logistic mixed effect model on our data.
>> The outcome variable is binary.
>> I got the error when I wanted to add a group factor to my model.
>>
>> My initial model is as below:
>>
>> model.a <- nlme(response~ 1/(1 + exp( -intercept- u0 - slope*TIME -
>> u1*TIME)),
>>                            + fixed=intercept+slope~1, random= u0+u1~1
>> |studentID,
>>                            + start=c(slope=.01, intercept=-1), data=log.data,
>> method='ML')
>>
>> This works fine on my data. But when i update it by adding a group factor
>> SKILLS,
>> I got the error message:
>> "Error in nlme.formula(response ~  1/(1 + exp( -intercept- u0 - slope*TIME
>> :
>>         starting values for the fixed component are not the correct length"
>>
>> The model is as below:
>> model.a <- nlme(response~ 1/(1 + exp( -intercept- u0 - slope*TIME -
>> u1*TIME)),
>>                            + fixed=intercept+slope ~ SKILLS, random= u0+u1~1
>> |studentID,
>>                            + start=c(slope=.01, intercept=-1), data=log.data,
>> method='ML')
>>
>> Does anybody see anything wrong with the "start" part of this model?
>>
>> Thanks a lot!!
>>
>>         [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> 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
>>
>
>
> --
> »ÆÈÙ¹ó
> Deparment of Sociology
> Fudan University
>
>

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595