Extended Poisson–Tweedie: Properties and regression models for count data

Abstract

We propose a new class of discrete generalized linear models based on the class of Poisson-Tweedie factorial dispersion models with variance of the form µ + φµ p , where µ is the mean, φ and p are the dispersion and Tweedie power parameters, respectively. The models are fitted by using an estimating function approach obtained by combining the quasi-score and Pearson estimating functions for estimation of the regression and dispersion parameters, respectively. This provides a flexible and efficient regression methodology for a comprehensive family of count models including Hermite, Neyman Type A, Pólya-Aeppli, negative binomial and Poisson-inverse Gaussian. The estimating function approach allows us to extend the Poisson-Tweedie distributions to deal with underdispersed count data by allowing negative values for the dispersion parameter φ. Furthermore, the Poisson-Tweedie family can automatically adapt to highly skewed count data with excessive zeros, without the need to introduce zero-inflated or hurdle components, by the simple estimation of the power parameter. Thus, the proposed models offer a unified framework to deal with under, equi, overdispersed, zero-inflated and heavy-tailed count data. The computational implementation of the proposed models is fast, relying only on a simple Newton scoring algorithm. Simulation studies showed that the estimating function approach provides unbiased and consistent estimators for both regression and dispersion parameters. We highlight the ability * Generalized linear models (GLMs) (Nelder and Wedderburn;1972) have been the main statistical tool for regression modelling of normal and non-normal data over the past four decades. The success enjoyed by the GLM framework comes from its ability to deal with a wide range of normal and non-normal data. GLMs are fitted by a simple and efficient Newton score algorithm relying only on second-moment assumptions for estimation and inference. Furthermore, the theoretical background for GLMs is well established in the class of dispersion models (Jørgensen; 1987, 1997) as a generalization of the exponential family of distributions. In particular, the Tweedie family of distributions plays an important role in the context of GLMs, since it encompasses many special cases including the normal, Poisson, non-central gamma, gamma and inverse Gaussian.In spite of the flexibility of the Tweedie family, the Poisson distribution is the only choice for the analysis of count data in the context of GLMs. For this reason, in practice there is probably an over-emphasis on the use of the Poisson distribution for count data. A well known limitation of the Poisson distribution is its mean and variance relationship, which implies that the variance equals the mean, referred to as equidispersion. In practice, however, count data can present other features, namely underdispersion (mean > variance) and overdispersion (mean < variance) that is often related to zero-inflation or a heavy tail. These departures can make the Poisson distribution unsuitable, or at lea...

Linking Land Use to Atlantic Salmon Production to Guide Recovery Planning

“…This package fits a Tweedie compound Poisson distribution for the response (log link), which is ideal for data that contains measured zeros as well as a continuous positive distribution of densities (Bonat et al. ). Coefficients are estimated using maximum likelihood based on the LaPlace approximation (Zhang ).…”

On Poisson–Tweedie mixtures

“…'s) which are hereinafter referred to as the Poisson-Tweedie mixtures. This family was considered, among many others, by Kokonendji et al (2004, Section 3), Jørgensen and Kokonendji (2016), and Bonat et al (2017). The Poisson-Tweedie mixtures are rigorously introduced by formula (3).…”

Modelagem estatística de dados pesqueiros

“…Para uma variável aleatória Z com distribuição Tweedie, escrevemos µ e Var(Y ) = µ + ϕµ p . Esse fato motiva a especificação de um modelo baseado apenas em suposições de momentos de segunda-ordem (Bonat et al, 2018). Sendo assim, a função de variância para esta família de distribuições é ϑ(µ; p) = µ + ϕµ p , em que modelos importantes para dados de contagem são casos particulares, como Hermite (p = 0),…”

“…As estimativas dos parâmetros de regressão são realizadas pelo método de quaseescore, e os parâmetros de dispersão pelas funções de estimação de Pearson. A formulação do modelo Poisson-Tweedie é exatamente a mesma daquela para os modelos quase-binomial e quase-Poisson no contexto de GLMs (Bonat et al, 2018). 4.3.…”