Poisson vs. negative binomial

227 views
Skip to first unread message

Neil Carter

unread,
Sep 7, 2011, 2:38:30 PM9/7/11
to unmarked
Hello,
I am trying to determine if I should use a Poisson or a negative
binomial model. What indicator should I use to determine if my data
is overdispersed? Does unmarked have a built-in method of determining
if the negative binomial is appropriate? For instance, these are the
results from the best-fit negative binomial model I have run on my
data so far. What exactly does the non-significant z-score tell me?
Appreciate any help.

pcount(formula = ~maj_prey + area_am_riv + block2 + block3 +
block4 ~ shape_am + pladj, data = tigerUMF, K = 22, mixture =
c("NB"))

Abundance:
Estimate SE z P(>|z|)
(Intercept) 1.746 0.305 5.72 1.05e-08
shape_am 0.867 0.210 4.13 3.65e-05
pladj 0.906 0.240 3.78 1.58e-04

Detection:
Estimate SE z P(>|z|)
(Intercept) -3.634 0.351 -10.34 4.55e-25
maj_prey 0.317 0.128 2.48 1.33e-02
area_am_riv -0.779 0.172 -4.53 5.88e-06
block2 -0.375 0.328 -1.14 2.53e-01
block3 -1.019 0.368 -2.77 5.61e-03
block4 -1.828 0.373 -4.90 9.56e-07

Dispersion:
Estimate SE z P(>|z|)
10.5 68.2 0.154 0.877

AIC: 861.418

Kery Marc

unread,
Sep 7, 2011, 3:21:42 PM9/7/11
to unma...@googlegroups.com
hi Neil,

people would normally look at the AIC from a model that assumes a Poisson and that which assumes a NB and choose that distribution which gives clearly a lower AIC (for instance, more than 2 AIC units); see for instance our early Nmix paper in Ecol.Appl. (Kery et al. 2005). Typically, the NB turns out to be favoured by AIC. Be aware that the NB can produce very high and sometimes unrealistic abundance estimates. If that happens, I would choose the Poisson. There is a paper on those unrealistic NB abundance estimates, about 2009 or 2010 somewhere in EcolApps or perhaps ConsBiol.

Regards -- Marc

______________________________________________________________

Marc Kéry
Tel. ++41 41 462 97 93
marc...@vogelwarte.ch
www.vogelwarte.ch

Swiss Ornithological Institute | Seerose 1 | CH-6204 Sempach | Switzerland
______________________________________________________________

***** Check out my recent book on Bayesian statistical modeling at www.mbr-pwrc.usgs.gov/pubanalysis/kerybook

***** Watch out for our new book (due Dec 2011): Kéry & Schaub, Bayesian Population Analysis using WinBUGS, Academic Press;
www.vogelwarte.ch/bpa

________________________________________
From: unma...@googlegroups.com [unma...@googlegroups.com] on behalf of Neil Carter [nhcar...@gmail.com]
Sent: 07 September 2011 20:38
To: unmarked
Subject: [unmarked] Poisson vs. negative binomial

Matthew Giovanni

unread,
Sep 7, 2011, 4:02:17 PM9/7/11
to unma...@googlegroups.com
Joseph et al. 2009 Ecological Applications

Neil Carter

unread,
Sep 7, 2011, 5:56:37 PM9/7/11
to unmarked
Thanks! This is very helpful.
--N

On Sep 7, 4:02 pm, Matthew Giovanni <matthewgiova...@gmail.com> wrote:
> Joseph et al. 2009 Ecological Applications
>
>
>
>
>
>
>
> On Wed, Sep 7, 2011 at 1:21 PM, Kery Marc <marc.k...@vogelwarte.ch> wrote:
> > hi Neil,
>
> > people would normally look at the AIC from a model that assumes a Poisson
> > and that which assumes a NB and choose that distribution which gives clearly
> > a lower AIC (for instance, more than 2 AIC units); see for instance our
> > early Nmix paper in Ecol.Appl. (Kery et al. 2005). Typically, the NB turns
> > out to be favoured by AIC. Be aware that the NB can produce very high and
> > sometimes unrealistic abundance estimates. If that happens, I would choose
> > the Poisson. There is a paper on those unrealistic NB abundance estimates,
> > about 2009 or 2010 somewhere in EcolApps or perhaps ConsBiol.
>
> > Regards  --  Marc
>
> > ______________________________________________________________
>
> > Marc Kéry
> > Tel. ++41 41 462 97 93
> > marc.k...@vogelwarte.ch
> >www.vogelwarte.ch
>
> > Swiss Ornithological Institute | Seerose 1 | CH-6204 Sempach | Switzerland
> > ______________________________________________________________
>
> > ***** Check out my recent book on Bayesian statistical modeling at
> >www.mbr-pwrc.usgs.gov/pubanalysis/kerybook
>
> > ***** Watch out for our new book (due Dec 2011): Kéry & Schaub, Bayesian
> > Population Analysis using WinBUGS, Academic Press;
> >www.vogelwarte.ch/bpa
>
> > ________________________________________
> > From: unma...@googlegroups.com [unma...@googlegroups.com] on behalf of
> > Neil Carter [nhcarte...@gmail.com]

Carlos

unread,
Sep 14, 2011, 8:24:10 AM9/14/11
to unmarked
Hi Marc!
Do you have more details about this citation?...author?
Thanks,
Carlos

Kristen Kirkby

unread,
Apr 6, 2012, 3:23:15 PM4/6/12
to unma...@googlegroups.com
It's seven months down the line now, but just in case you're still curious, I think it might be:

Modeling Abundance Using N-Mixture Models: The Importance of Considering Ecological Mechanisms
Liana N. Joseph, Ché Elkin, Tara G. Martin and Hugh P. Possingham
Ecological Applications , Vol. 19, No. 3 (Apr., 2009), pp. 631-642


Kristen
Reply all
Reply to author
Forward
0 new messages