Estimação de escores binomiais correlacionados: uma aplicação em Credit Scoring

Abstract

For the most part of modelings in the credit risk area, the most widely used model is the credit scoring, and as the main statistical technique, the binary logistic regression, used to determine whether a customer is a good or bad payer. In this academic work an alternative methodology is proposed, where the estimative is formed based on the scores obtained by customers; this means the response follows a binomial distribution. In this modeling the combined estimate of scores of various products used by customers is included, considering the correlation between these scores.