R help - May 2010 - Estimating theta for negative binomial model

Dear List,

I am trying to do model averaging for a negative binomial model using the
package AICcmodavg.  I need to use glm() since the package does not accept
glm.nb() models.  I can get glm() to work if I first run glm.nb and take theta
from that model, but is there a simpler way to estimate theta for the glm model?
The two models are:

mod.nb<-glm.nb(mantas~site,data=mydata)
mod.glm<-glm(mantas~site,data=mydata, family=negative.binomial(mod.nb$theta))

How else can I get theta for the family=negative.binomial(theta=???) 

Thanks!

Tim



Tim Clark
Department of Zoology 
University of Hawaii

On Mon, 3 May 2010, Tim Clark wrote:
> Dear List,
>
> I am trying to do model averaging for a negative binomial model 
> using the package AICcmodavg.  I need to use glm() since the package 
> does not accept glm.nb() models.  I can get glm() to work if I first 
> run glm.nb and take theta from that model, but is there a simpler 
> way to estimate theta for the glm model?  The two models are:
>
> mod.nb<-glm.nb(mantas~site,data=mydata)
> mod.glm<-glm(mantas~site,data=mydata,
family=negative.binomial(mod.nb$theta))
>
> How else can I get theta for the family=negative.binomial(theta=???)
library(MASS) is missing here -- do give credit where credit is due.
It contains functions

glm.convert
theta.ml

and the first does what you seem to want and the second answers the 
actual question you asked.
> Thanks!
>
> Tim
>
> Tim Clark
> Department of Zoology
> University of Hawaii
-- 
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

Brian,

glm.convert() does exactly what I am wanting.  Thanks for putting it in MASS. 
Sorry about not giving credit.  It is a great package and I keep finding
functions that make statistical programming easy!

Aloha,

Tim


Tim Clark
Department of Zoology 
University of Hawaii


--- On Mon, 5/3/10, Prof Brian Ripley <ripley at stats.ox.ac.uk> wrote:
> From: Prof Brian Ripley <ripley at stats.ox.ac.uk>
> Subject: Re: [R] Estimating theta for negative binomial model
> To: "Tim Clark" <mudiver1200 at yahoo.com>
> Cc: r-help at r-project.org
> Date: Monday, May 3, 2010, 7:50 PM
> On Mon, 3 May 2010, Tim Clark wrote:
> 
> > Dear List,
> > 
> > I am trying to do model averaging for a negative
> binomial model using the package AICcmodavg.? I need to
> use glm() since the package does not accept glm.nb()
> models.? I can get glm() to work if I first run glm.nb
> and take theta from that model, but is there a simpler way
> to estimate theta for the glm model?? The two models
> are:
> > 
> > mod.nb<-glm.nb(mantas~site,data=mydata)
> > mod.glm<-glm(mantas~site,data=mydata,
> family=negative.binomial(mod.nb$theta))
> > 
> > How else can I get theta for the
> family=negative.binomial(theta=???)
> 
> library(MASS) is missing here -- do give credit where
> credit is due.
> It contains functions
> 
> glm.convert
> theta.ml
> 
> and the first does what you seem to want and the second
> answers the actual question you asked.
> 
> > Thanks!
> > 
> > Tim
> > 
> > Tim Clark
> > Department of Zoology
> > University of Hawaii
> 
> -- 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
>