R help - Jul 2011 - Deviance of zeroinfl/hurdle models

Dear list, I'm wondering if anyone can help me calculate the deviance
of either a zeroinfl or hurdle model from package pscl?
Even if someone could point me to the correct formula for calculating
the deviance, I could do the rest on my own.

I am trying to calculate a pseudo-R-squared measure based on the
R^{2}_{DEV} of [1], so I need to be able to calculate the deviance of
the full and null models. Does anyone have any suggestions?
Alternatively, does anyone have a suggestion for a better measure to
report (I'm aware that R^2 measures aren't really appropriate here),
preferably something that is easy enough to program or compute using
existing packages...

Thanks in advance,

Carson

[1] Cameron, A.C., Windmeijer, F.A.G., 1996. R^2 measures for count
data regression models with applications to health-care utilization.
J. Bus. Econom. Statist. 14, 209?220


-- 
Carson J. Q. Farmer
ISSP Doctoral Fellow
National Centre for Geocomputation
National University of Ireland, Maynooth,
http://www.carsonfarmer.com/

Carson Farmer <carson.farmer <at> gmail.com> writes:
> 
> Dear list, I'm wondering if anyone can help me calculate the deviance
> of either a zeroinfl or hurdle model from package pscl?
> Even if someone could point me to the correct formula for calculating
> the deviance, I could do the rest on my own.

    What about 

library(pscl)
example(hurdle)
-2*logLik(fm_hnb2) 

 ?

bbolker wrote:
> 
>     What about 
> 
> library(pscl)
> example(hurdle)
> -2*logLik(fm_hnb2) 
> 
>  ?
> 
Should this not be -2*(logLik(M) - logLik(S)) where M is the current model
and S is the saturated or full model? In which case, is it safe to assume
that L(S) = 0? Or have I got my deviances crossed?

Carson

-----
Carson J. Q. Farmer
Doctoral Fellow
National Centre for Geocomputation
National University of Ireland, Maynooth
--
View this message in context:
http://r.789695.n4.nabble.com/Deviance-of-zeroinfl-hurdle-models-tp3663196p3679780.html
Sent from the R help mailing list archive at Nabble.com.