PUB - Publikationen an der Universität Bielefeld

Procedures for determining the feasibility of lower and upper bounds on Euclidean distances of fixed dimension play a central role in the analysis of many kinds of scientific data. Shown in this paper is how results from graph optimization theory can be used to solve the feasibility problem in one dimension, subject to the condition that the order of the points along the real line is known. The solution is used to derive a PSPACE, O(n3 . n!)-time sequential algorithm for finding one-dimensional representations subject to arbitrary distance (and order) constraints. The wider applicability of these results in measurement theory is discussed, in particular, Roy's elegant proofs of the classical representation theorems for interval orders and semiorders, and they are used to obtain a new representation theorem for a ternary relation called epsilon-collinearity.