Análise geométrica de dados através de análise de correspondência múltipla

Abstract

In this work we applied the multivariate statistical technique, called Multiple Correspondence Analysis (MCA), following an approach based on geometric analysis of data from a corneal transplantation database, where the lines represent the corneas collected from donors after death and the columns represent the different categories of variables that characterize the corneas and donors. The information was obtained from medical records. The database comprises a set of real data with 264 individuals and 8 variables, obtained from the Eye Bank of Joa ̃o Penido Hospital. As Correspondence Analysis (CA) is de- veloped using categorical variables or continuous variables which have been categorized, and some variables are numerical in nature, we used a method to group them into ordi- nal categories corresponding to quartiles, although the technique treat them as nominal categories. The goal is to show geometrically the rows and columns of the data table in a low-dimensional space, so that the proximity in space indicating associations, similarity or dissimilarity of categories and individuals. Through the graphs of the ACM, we can analyze separately the variables, classes, individuals, and both individuals and categories. The implementation of the MCA was performed using the free software R.