Distribuições de probabilidade para análise de dados de contagem de parasitos em bovinos

• Main Author: Andresa Eva Melo de Araujo
• Publication Date: 2019
• Format: Master thesis
• Language: por
• Source: Repositório Institucional da UFMG
• Download full: http://hdl.handle.net/1843/32485
• topic: Inferência Bayesiana; Poisson; Binomial Negativa

Traits of resistance against parasites are considered as selection criteria in genetic improvement programs of cattle due to the economic impact associated with parasitism. In general, resistance traits are studied through count records of ticks, parasite eggs and oocysts and those counts are analyzed through different probability distributions. We aimed to identify whether there would be any differences in the estimation of variance and in the selection and ranking of animals for parasite counts between Poisson (P), zero-inflated Poisson (ZIP), negative binomial (NB), and zero-inflated negative binomial (ZINB) distributions. We analyzed 7,858 tick counts and 7,144 counts of both gastrointestinal nematodes (GIN) and Eimeria spp. through Bayesian inference. The distribution which best fitted the data was Poisson. Estimates of permanent environmental and contemporary group variances for tick and Eimeria spp. counts, and also of additive genetic and permanent environmental variances for NGI counts were similar between P and ZIP distributions. Between NB and ZINB distributions, all variances showed some similarity. In the animal selection, the similarity of individuals equally selected between P and ZIP showed a variation of 50-94% for ticks, 67-85% for GIN and 29-50% for Eimeria spp. Between NB e ZINB, this similarity was always near 100%. Neither NB nor ZINB showed any differences in animal ordering at the bull ranking, although in the other distributions the observed changes were not significant. Our results indicate that there are differences in the estimates of variance obtained between P, ZIP, NB and ZINB distributions. For practical purposes, the utilization of P, NB and ZINB distributions will not interfere in animal selection and ranking.