STATIS-CNS2: an exploratory analysis technique for qualitative non-symmetric data

Abstract

The problem raised in this research focuses on the search for a methodology, based on STATIS, that allows comparing and
simultaneously explaining the influence of an explanatory qualitative variable x as determinant of a criterion categorical variable y
on H occasions. To this end, a Frobenius scalar product is defined between the objects, which allows us to conceptualize a statistical distance between objects, based on the Goodman-Kruskal statistic τ.
An application of the proposed technique on a set of real data from the financial field is also presented made up of 4 blocks of
paired data, where each block contains the measurement of two qualitative variables on a sample of 20316 individuals, with the
purpose of measuring the impact it has on the information emanating from the Colombian credit bureaus the segmentation used for the calculation of provisions both in local accounting and accounting under IFRS9 regulations.