Análise de dados de gastroenterite hemorrágica canina para identificar fatores de risco por regressão logística

Abstract

This paper presents a study of how to fit a logistic regression model to predict the death probability of dogs with hemorrhagic gastroenteritis. A logistic model is recommended to treat dichotomic variables in Coorte study. Using a census procedure from 1992 to 1999 four veterinary clinic in Lavras, MG, registered 176 infected animals. The variables of the model have been chosen to be sex, age internment days rates and number of clinical treatments by the or Fisher's exact test. The parameters were estimated by the maximum likelihood method. The results showed that if the infected dogs were clinically treated only once then the animals older than six months had their mortality chances 15.45 times (P<0.05) larger than those younger than six months. If the infected animals younger than six months were clinically treated only once then their mortality chances were 20.251 (P<0.05) higher than if they had received two to seven medical treatments.