R help - Apr 2004 - Code for quasi-likelihood binomial estimation

Hello,

Has anyone written up code to estimate for example a simple logit using 
quasi-likelihood?  I know that glm() already does this, but I'd like to do 
some tinkering with the variance function beyond what glm() allows.  I've 
scanned online sources and everyone seems to use glm().  Will take a crack 
at it if necessary but have zero experience w/q-likelihood and not that much 
with coding; thought I'd check here first so as not to reinvent the wheel, 
perhaps learn a bit from what others have already done.

Thanks,
GF

_________________________________________________________________
Stop worrying about overloading your inbox - get MSN Hotmail Extra Storage!

Gordon Fredericks wrote:
> Hello,
> 
> Has anyone written up code to estimate for example a simple logit using 
> quasi-likelihood?  I know that glm() already does this, but I'd like to
> do some tinkering with the variance function beyond what glm() allows.  
> I've scanned online sources and everyone seems to use glm().  Will take
> a crack at it if necessary but have zero experience w/q-likelihood and 
> not that much with coding; thought I'd check here first so as not to 
> reinvent the wheel, perhaps learn a bit from what others have already done.
> 
> Thanks,
> GF
> 
See ?quasibinomial.

?glm would have pointed you to ?family which would have shown you the 
quasi family.

--sundar

On Fri, 30 Apr 2004, Gordon Fredericks wrote:
> Hello,
>
> Has anyone written up code to estimate for example a simple logit using
> quasi-likelihood?  I know that glm() already does this, but I'd like to
do
> some tinkering with the variance function beyond what glm() allows.
This should just be a matter of writing a family() object for glm.  I'd
recommend looking at the negative binomial example in MASS (the book as
well as the package).

	-thomas