Bayesian multidimensional scaling for the estimation of a Minkowski exponent

Area of Science:

• Psychology
• Cognitive Science
• Mathematical Psychology

Background:

• The Minkowski property of psychological space is a key research area.
• Current methods for determining the Minkowski exponent, like minimizing stress in multidimensional scaling, suffer from arbitrariness.
• A recent Bayesian approach shows promise but was designed for individual data, leaving its applicability to averaged data uncertain.

Purpose of the Study:

• To evaluate the applicability of an existing Bayesian method to averaged psychological data.
• To develop a novel method for accurately estimating the Minkowski exponent from averaged or single-subject data in psychological research.

Main Methods:

• A simulation study was conducted to test the existing Bayesian method on averaged data.
• A new Bayesian multidimensional scaling method was developed, extending the Euclidean approach to the Minkowski metric.
• A second simulation study validated the performance of the proposed method.

Main Results:

• The existing Bayesian method failed to accurately recover the true Minkowski exponent when applied to averaged data.
• The newly proposed Bayesian multidimensional scaling method successfully recovered the true Minkowski exponent in simulations.
• The developed method offers a reliable solution for analyzing averaged psychological data.

Conclusions:

• The existing Bayesian approach is not suitable for averaged psychological data.
• The proposed extension of Bayesian multidimensional scaling to the Minkowski metric provides an effective solution for estimating psychological space exponents.
• This new method addresses a critical gap in analyzing common psychological and behavioral science data.