Uma nova abordagem para análise de dependência bivariada

Abstract

In this dissertation we describe and implement procedures for nonparametric estimation of copulas and Sibuya function, and also procedures for bivariate analysis of dependence based on the behavior of their contours plot. Besisdes, we describe and implement the chiplot procedure and as well as a procedure for analising bivariate dependence in presence of censoring in the sample. Particularly, we propose a way to use it in a local correlation analysis. The performance of the proposed procedures are illustrated and evaluated in cases of very simple correlation, but also in a more complex correlation schemes.