Modelos lineares generalizados multinomiais univariado e bivariado

Abstract

This work presents the basical theorics fundaments of the generalized linear models e the caracteristiscs of the univariate multinomial and binomial logistic regression analysis. Practical aplications of binomial regression and of multinomial were perfomed. The aplication were related the evaluation of effectiveness of an herbicide for the control of a certain weed; of the effectiveness of the radiation level for the house flies control; and for the evaluation of the degree of retinopathy as a function of the individuals' diabetes and other condition. The code of the implementation of the analitycals procedures, run by the softwares R, MATLAB e SAS are avaliable. Also are descrited three bivariate multinomial logistic reggression models: fixed effects conditional, margimal and mixed effects. Applications for each models are presented. The applications are about: the degree of retinopathy avaliation in both eyes of patientsas a function of diabetics and other individual features; the severity of rheumatoid arthritis as a function of the patients status. The code of the implementation of the analitycals procedures, run by the softwares R, MATLAB e SAS are avaliable.