The effects of incorrect specification of the Minkowski metric: A simulation experiment

Abstract

This is the second article in a two part investigation into the consequences of incorrectly specifying the value of the Minkowski metric in multidimensional scaling studies. The first article, which appeared in the Summer 1974 issue of this journal, established five theorems which hold if the axes of the true and derived solution spaces have the same orientations. It was further shown that four of these theorems hold only for the special Euclidean case when the axis orientations of the true and derived solution spaces are different. This second article extends these results by investigating the sensitivity of the TORSCA nonmetric multidimensional scaling algorithm to incorrect specification of the Minkowski metric and the ability of Krukal's “stress” measure to detect such incorrect specification.