Models for zero-inflated and overdispersed proportion data – A bayesian approach

Adriano Ferreti BORGATTO[1]

Clarice Garcia Borges DEMÉTRIO[2]

Roseli Aparecida LEANDRO2

§    ABSTRACT: In general, standard binomial regression models do not fit well to proportion data from biological control assays, mainly when there is excess of zeros and overdispersion. In this study, beta-binomial and zero-inflated beta-binomial models are applied to a data set obtained from a biological control assay to produce parasitized eggs to control Diatraea saccharalis, a common pest in sugar cane. A parasite (Trichogramma galloi) was put to parasitize 128 eggs of Anagasta kuehniella, an economically suitable alternative host, with a variable number of female parasites (2, 4, 8,..., 128), each with 10 replicates in a completely randomized experiment. A Bayesian procedure was formulated using a simulation technique (Metropolis Hastings) for estimation of the parameters of interest. The convergence of the Markov Chain generated was monitored by visualization of the trace plot and using Raftery & Lewis and Heidelberg & Welch diagnoses presented in module CODA of software R.

§    KEYWORDS: Generalized linear models; Bayesian analysis; binomial model; overdispersion; excess of zeros; MCMC method.

 



[1]Departamento de Informática e Estatística, Universidade Federal de Santa Catarina -  UFSC, CEP 88040-900, Florianópolis, SC, Brasil. E-mail: borgatto@inf.ufsc.br

[2]Departamento de Ciências Exatas, Universidade de São Paulo, campus de Piracicaba – ESALQ/USP, CEP 13418-900, Piracicaba, SP, Brasil. E-mail: clarice@carpa.ciagri.usp.br / raleandr@carpa.ciagri.usp.br