Row and column matrices in multiple correspondence analysis with ordered categorical and dichotomous variables

Abstract

In multiple correspondence analysis, whenever the number of variables exceeds the number of observations, row matrix should be used, but if the number of variables is less than the number of observations column matrix is the suitable procedure to follow. One of the following matrices (rows, columns) leads to loss of information that can be found by the other method, therefore, this paper developed a proposal to overcome this problem, which is: to find a shortcut method allowing the use of the results of one matrix to obtain the results of the other matrix. Taking advantage of all information available, the phenomenon was studied. Some of these results are: Eigenvectors, factor loadings and factor scores based on ordered categorical and dichotomous data. This method is illustrated by using a real data set. Results were obtained by using Minitab program. As a result, it is possible to shortcut transformation between the results of row and column matrices depending on factor loadings and factor scores of the row and column matrices.