Skip to content

package for zero-inflated poisson models with temporal correlation?

2 messages · Trina Schneider Bayard, David Atkins

Original
Trina Schneider Bayard
# 
Dear all,

After looking through the archives I'm pretty certain I'm on the right
list but please let me know if this question is best posted elsewhere:

Does anyone know of an R package that is able to perform zero-inflated
Poisson/negative binomial models that incorporate temporal correlation?

My data consist of weekly counts (12) of various bird behaviors taken at
29 points, along with various explanatory variables related to local
reproductive activity and environmental conditions.

So far, I have been able to
1) model the counts of bird behaviors as a function of the explanatory
variables while incorporating the temporal correlation associated with my
data collection methods using the ?geepack? library (thus ignoring
zero-inflation)
2) model the zero-inflated counts of bird behaviors as a function of the
explanatory variables using the ?pscl? library (thus ignoring temporal
correlation)
....but have not been able to find a function that incorporates both
conditions.

I?d rather not take the average of the counts across all 12 weeks as I am
interested in how specific bird behaviors change according to both local
reproductive activities and the progression of the breeding season.

Thanks in advance for your help,

Trina
David Atkins
# 
Hi Trina--

In short: no, I'm not aware of a package that can do zero-inflated 
(mixed/GEE) models with autocorrelated errors.

However, the MCMCglmm package can fit mixed-effects zero-inflated 
models, and a random-effect for time implies some autocorrelation. 
Thus, one option would be to use MCMCglmm and explore whether there 
appears to be residual autocorrelation, and compare this to your results 
using GEE and zeroinfl() in pscl.

Finally, Zuur et al. show BUGS code for mixed-effects Poisson with 
autocorrelated error in their book, Mixed Effects Models and Extensions 
in Ecology with R.

For what it's worth.

cheers, Dave

-- 
Dave Atkins, PhD
Research Associate Professor
Department of Psychiatry and Behavioral Science
University of Washington
datkins at u.washington.edu

Center for the Study of Health and Risk Behaviors (CSHRB)		
1100 NE 45th Street, Suite 300 	
Seattle, WA  98105 	
206-616-3879 	
http://depts.washington.edu/cshrb/
(Mon-Wed)	

Center for Healthcare Improvement, for Addictions, Mental Illness,
   Medically Vulnerable Populations (CHAMMP)
325 9th Avenue, 2HH-15
Box 359911
Seattle, WA 98104?
206-897-4210
http://www.chammp.org
(Thurs)


Dear all,

After looking through the archives I'm pretty certain I'm on the right
list but please let me know if this question is best posted elsewhere:

Does anyone know of an R package that is able to perform zero-inflated
Poisson/negative binomial models that incorporate temporal correlation?

My data consist of weekly counts (12) of various bird behaviors taken at
29 points, along with various explanatory variables related to local
reproductive activity and environmental conditions.

So far, I have been able to
1) model the counts of bird behaviors as a function of the explanatory
variables while incorporating the temporal correlation associated with my
data collection methods using the ?geepack? library (thus ignoring
zero-inflation)
2) model the zero-inflated counts of bird behaviors as a function of the
explanatory variables using the ?pscl? library (thus ignoring temporal
correlation)
....but have not been able to find a function that incorporates both
conditions.

I?d rather not take the average of the counts across all 12 weeks as I am
interested in how specific bird behaviors change according to both local
reproductive activities and the progression of the breeding season.

Thanks in advance for your help,

Trina