Some days ago I asked whether someone has S-PLUS (or R) code for the
method proposed by
Efron B. (1986).
Double exponential families and their use in generalized linear regression.
JASA 83:709-721.
I got the following two results.
The website for the code dglm mentioned by Hugh Jones concerning an estimation
method
proposed by Smyth and Verbyla (1999) seems to be now:
http://www.statsci.org/s/dglm.html
http://www.statsci.org/s/digammaf.html
http://www.statsci.org/s/polygamm.html
http://www.statsci.org/s/tweedief.html
Thank you,
Andreas
======================================
HUGH JONES wrote:
I recently came across a paper by Smyth, G.K., and Verbyla, A.P. (1999).
Adjusted likelihood methods for modelling dispersion in Generalized Linear Models. Environmetrics 10: 695-709.
These authors consider double generalized models for modelling both mean and dispersion simultaneously.
Is this what Efron's paper is about? They wrote two functions for applying double GLMs called dglm( )
and a new error family tweedie based on work by Tweedie 1984 (An index which distinguishes
between some important exponential families. In, 'Statistics: Applications and New Directions' (eds Ghosh & Roy))
These are available from http://www.maths.uq.edu.au/~gks/s/
======================================
JEFF SIMONOFF wrote:
James Lindsey, in his 1995 book _Modelling Frequency and Count Data_
(Clarendon Press) shows how distributions of this type (without
covariates, however) can be easily fit using standard loglinear (Poisson
regression) software - see pages 125-132. I would be interested to hear
about ready-made software for the regression (covariate) case, however.
=======================================
Andreas Christmann
Universitaet Dortmund
HRZ
44221 Dortmund
GERMANY
email A.Christmann@hrz.uni-dortmund.de
phone (049)-231-755-2763
fax (049)-231-755-2731
=======================================
|