Comparação e aplicação de modelos de regressão binária na retenção de capacetes de motociclistas

Abstract

Helmet is a security equipment of mandatory use for Brazilian motorcyclists. It is compost of an outer wrapper and a retention system. The usefulness of such a system is keeping the helmet on the head. The helmet can be used in the correct or incorrect way. If used incorrectly, the helmet can be more likely ejected, increasing the mortality of the motorcyclists. The main goal of this paper is to identify factors associated with the incorrect use of the helmet. We evaluated the helmet type, coupling type, sex and position on the motorcycle. The data was collected in Curitiba, PR, Brazil. We fitted generalized linear models using the Bernoulli distribution and compared the fit of eight link functions: logit, probit, complement log-log, Cauchy, Aranda-Ordaz, Weibull, Prentice and Studel. For our data, the Cauchy link function presented the best fit. The results showed that gender is not statistically signicant, however, helmet type, coupling type and position on the motorcycle are important factors to determine the probability of correct use of the helmet.