×

A computational short-cut for INDSCAL with orthonormality constraints on positive semi-definite matrices for low rank. (English) Zbl 0715.65120

When INDSCAL with orthonormality constraints (INDORT) is to be applied to a set of very large similarity matrices, one is confronted with huge computational problems. When those very large matrices are positive semi- definite (p.s.d.) matrices of low rank, computations can be facilitated to a large extent. The present paper describes a simplified algorithm for the INDORT analysis of such matrices. One of the applications of INDORT on low-rank p.s.d. matrices is that of INDORT on a set of quantification matrices for qualitative and/or quantitative variables. The implications of using the simplified algorithm for this INDORT analysis are worked out. It turns out that INDORT for qualitative data can be applied when one only has the total bivariate contingency table for all variables.

MSC:

65C99 Probabilistic methods, stochastic differential equations
62H17 Contingency tables
PDFBibTeX XMLCite