Quasi-Poisson vs. negative binomial regression: how should we model overdispersed count data?

National Marine Mammal Laboratory, Alaska Fisheries Science Center, National Marine Fisheries Service, 7600 Sand Point Way NE, Building 4, Seattle, Washington 98115-6349, USA.
Ecology (Impact Factor: 5). 12/2007; 88(11):2766-72. DOI: 10.1890/07-0043.1
Source: PubMed

ABSTRACT Quasi-Poisson and negative binomial regression models have equal numbers of parameters, and either could be used for overdispersed count data. While they often give similar results, there can be striking differences in estimating the effects of covariates. We explain when and why such differences occur. The variance of a quasi-Poisson model is a linear function of the mean while the variance of a negative binomial model is a quadratic function of the mean. These variance relationships affect the weights in the iteratively weighted least-squares algorithm of fitting models to data. Because the variance is a function of the mean, large and small counts get weighted differently in quasi-Poisson and negative binomial regression. We provide an example using harbor seal counts from aerial surveys. These counts are affected by date, time of day, and time relative to low tide. We present results on a data set that showed a dramatic difference on estimating abundance of harbor seals when using quasi-Poisson vs. negative binomial regression. This difference is described and explained in light of the different weighting used in each regression method. A general understanding of weighting can help ecologists choose between these two methods.

  • Source
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Top-down control of prey by predators are magnified in productive ecosystems due to higher sustenance of prey communities. In soil micro-arthropod food webs, plant communities regulate the availability of basal resources like soil microbial biomass. Mixed plant communities are often associated with higher microbial biomass than monocultures. Therefore, top-down control is expected to be higher in soil food webs of mixed plant communities. Moreover, higher predator densities can increase the suppression of prey, which can induce interactive effects between predator densities and plant community composition on prey populations. Here, we tested the effects of predator density (predatory mites) on prey populations (Collembola) in monoculture and mixed plant communities. We hypothesized that top-down control would increase with predator density but only in the mixed plant community. Our results revealed two contrasting patterns of top-down control: stronger top-down control of prey communities in the mixed plant community, but weaker top-down control in plant monocultures in high predator density treatments. As expected, higher microbial community biomass in the mixed plant community sustained sufficiently high prey populations to support high predator density. Our results highlight the roles of plant community composition and predator densities in regulating top-down control of prey in soil food webs. T rophic interactions in food webs are observed as a combination of top-down and bottom-up control among consumers and their resources that principally regulate population dynamics in ecosystems 1,2. In tri-trophic interactions, either stronger top-down control (predator-induced control) or stronger bottom-up control (plant-induced control) of herbivore/detritivore populations have commonly been observed 3,4
    Scientific Reports 03/2015; 5. DOI:10.1038/srep09134 · 5.08 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The manipulation of habitat through thinning and logging activities may have cascading effects on fauna because of direct and indirect changes to key habitat features. However, the combined effect of thinning and logging on fauna has not been investigated for Australian forests, despite the widespread use of thinning as a silvicultural management tool in selectively logged forests, and the emerging interest in thinning for biofuels. We therefore surveyed reptiles, birds and key habitat variables at sites selected to sample four management classes reflecting categories of thinning and/or logging activity in cypress pine Callitris glaucophylla forests of the Brigalow Belt bioregion of Australia. Habitat structure and assemblages of reptile and bird species were distinct across the four management classes, reflecting long-term cumulative impacts of small- and/or large-diameter tree removal in a dynamic system, but the response of various functional groups of species was mixed. Recovery of some habitat elements (e.g. small trees) to a state that maintains densities of most fauna species appears to be relatively rapid, but depends on the combined effect of thinning and logging. Other habitat elements, such as large cypress trees which were reduced by logging and unaffected by thinning, require longer time frames to mature and therefore will be critical resources to maintain during future management activities in cypress forests. Overall, it appears that thinning activities affect fauna and therefore must be considered when making decisions about forest management. We suggest management of production forests aims for a mosaic of thinning and/or logging combinations across the landscape, but emphasise the importance of retaining or restoring unthinned and unlogged areas.
    Biological Conservation 06/2015; 186. DOI:10.1016/j.biocon.2015.03.009 · 4.04 Impact Factor

Full-text (3 Sources)

Available from
Dec 3, 2014