Anthropometric accommodation percentage comparison for univariate and multivariate representative Mexican hand models

Abstract

BACKGROUND: Adopting awkward postures at work has a great impact on productivity and work-related musculoskeletal disorders. Considering anthropometric data in the design of products and workplaces can diminish this impact. The traditional univariate-percentile-approach is one of the most implemented in the anthropometric analysis, even though it has proved limitations in comparison with multivariate-approaches. OBJECTIVE: To develop univariate and multivariate hand models considering four anthropometric dimensions, and to theoretically compare the univariate and multivariate accommodation percentages. METHODS: Univariate percentile models corresponding to the database of real subject nearest-neighbors to the 5th and 95th percentiles were obtained for the male and female population. Two multivariate approaches were implemented on the central 90% of both populations: 2D principal component analysis and archetypal analysis. The accommodation percentage for each family of models was obtained based on the population that simultaneously fit all the anthropometric dimensions. The goodness-of-fit and McNemar’s tests were performed to statistically analyze the accommodation percentages. RESULTS: Eight human hand models were obtained via Principal Component Analysis while two, three, four, and eight Archetypal Analysis models (male-population) and two, three, six, and eight Archetypal Analysis models (female-population) were selected after a root-sum-of-squares analysis for k = 1, ...  ,10 archetypes. CONCLUSION: The results showed that the Principal Component Analysis models obtained a higher accommodation level, followed by the Archetypal Analysis and percentile models (male population). In the case of the female population, models obtained by multivariate-Archetypal Analysis (k = 8) obtained a higher accommodation percentage.