[R-sig-ME] Negative binomial mixed effects modeling

[David Duffy]

On Wed, 12 May 2010, Ben Bolker wrote:

> David Duffy wrote:
>> On Tue, 11 May 2010, David Zajanc wrote:
>> If you have only a "single level" of random effects eg simple clusters,
>> then gamm() [mgcv] or spm() [SemiPar].
>
>  I don't see how this works -- gamm() will do the family options from
> glm (including negative.binomial() from MASS with a *fixed*
> overdispersion/theta/k parameter, but not? estimating the overdispersion
> as part of the fitting process?), spm() doesn't seem to have a NB option?
>
>  I would love to be wrong on this and corrected on it.

No you are right.  I don't regard having to iterate mgcv's gamm() "by 
hand" over theta to be a showstopper, though.  My memory was just wrong 
about spm().


Cheers, David.


-- 
| David Duffy (MBBS PhD)                                         ,-_|\
| email: davidD at qimr.edu.au  ph: INT+61+7+3362-0217 fax: -0101  /     *
| Epidemiology Unit, Queensland Institute of Medical Research   \_,-._/
| 300 Herston Rd, Brisbane, Queensland 4029, Australia  GPG 4D0B994A v