Re: [R] are hurdle logit-poisson model and posson model nested?

2007-09-24 Thread Achim Zeileis 
On Sat, 15 Sep 2007, Wensui Liu wrote:

 Dear Listers,
 I have a general statistical question. Are hurdle logit-poisson model
 and posson model nested?

No, I don't think so. The logit hurdle is equivalent to a geometric hurdle 
(i.e., logit and right-censored geometric distribution imply the same 
likelihood and thus the same estimates), so a hurdle logit-geometric model 
is nested withing a geometric model.
If you want a model nested in poisson, you can use a hurdle 
poisson-poisson model, i.e., a right-censored poisson for the zero hurdle 
and a left-truncated poisson for the counts. A test for presence of the 
hurdle is then a test that all parameters are equal. The hurdle() function 
in pscl can fit these models and hurdletest() can test for the presence 
of the hurdle (given that the same distribution has been used for the zero 
hurdle and the counts).

Best,
Z

 Thank you so much?

 __
 R-help@r-project.org mailing list
 https://stat.ethz.ch/mailman/listinfo/r-help
 PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
 and provide commented, minimal, self-contained, reproducible code.



__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

Re: [R] are hurdle logit-poisson model and posson model nested?

2007-09-17 Thread bbolker 



Wensui Liu wrote:
 
 Dear Listers,
 I have a general statistical question. Are hurdle logit-poisson model
 and posson model nested?
 

  You might have to give us a little more detail.  On first glance, my
impression
is that the Poisson model is _not_ nested in the hurdle-Poisson, because the
latter is truncated; what parameter(s) would you set to a fixed value to get
a regular Poisson?  (This is not true, on the other hand, of a zero-inflated
Poisson,
where you could set the zero-inflation parameter to zero to get a Poisson.)
[On the other hand, you would still have a parameter on the boundary of
its feasible space, which would make inference via LRT questionable.)

  good luck
Ben Bolker

-- 
View this message in context: 
http://www.nabble.com/are-hurdle-logit-poisson-model-and-posson-model-nested--tf4449782.html#a12745585
Sent from the R help mailing list archive at Nabble.com.

__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.