PROPOSAL FOR THE ANALYSIS OF BINARY CORRESPONDENCES MULTIPLE TIMES: AN ADAPTATION OF THE STATIS METHODOLOGY

In this research, the theoretical and applied foundations of an analytical proposal based on an adaptation of the STATIS methodology is formally presented. It makes use of the binary correspondence analysis to address the statistical treatment of K blocks of information generated by characterizing the same individuals, using the same two categorical variables, on different occasions. The proposed procedure, which has been designed and validated in simulated environments and implemented on real data, performs an analysis of the variability of the row or column profiles, of the contingency tables within each block, following the classic approach that defines the BCA, and also, fundamentally, performs a comparative analysis of the changes that occur in the variability of the profiles between the different blocks, throughout the different occasions. The results of these analysis have been summarized on a graphic device, an scatter diagram, in which the points in the plane are determined by the blocks, on which it is possible to visualize the magnitude that describes the measure of variability within each block in terms of the length of the vector that joins the block with the origin of the representation space. It has also been found that the smaller the distance between the information structures corresponding to two different blocks, the stronger the association between the variables that identify those blocks is greater, and vice versa for small distances between blocks. An application is presented on data from the Colombian Institute for the Evaluation of Education (ICFES) in which the educational units of the departments are classified according to the academic results obtained by the students in the Saber 11 test, for the periods 2015, 2016, and 2017.