Análisis discriminante no métrico y regresión logística en el problema de clasificación

Abstract

A study of comparison between non-metric discriminant analysis and logistic regression when the classification system has more than two clusters is shown, taking into account normal and non-normal distributions under different sample sizes. This process was carried out by means of a simulation study, evaluating the two procedures by the rate of misclassification. This study is derived from a research project and concluded that symmetrical distributions under the two procedures are similar in terms of the rate of misclassification and non-symmetrical distributions under the logistic regression performs better than the non metric discriminant analysis.