Regression models for predicting anthropometric measurements of students needed for ergonomics school furniture design

Abstract

This paper deals with the development of models that make use of some easy-to-measure anthropometric dimensions to predict difficult-to-measure dimensions required for ergonomic design of school furniture. A total of 143 students aged between 16 and 18 years from eight public secondary schools in Ogbomoso, Nigeria participated in the study. In a bid to avoid model complexity, a brute-force search implemented in Adaptive Neuro-Fuzzy Inference System (ANFIS) was employed to select the two most influential of the five input measurements. This search was separately conducted for each of the output measurements. Regression models were developed from the collected anthropometric data. Also, the predictive performance of these models was examined using ANOVA. The best models are the ones with no/non-significant lack of fit and highest coefficient of determinations. Nine out of the 12 developed models exhibit nonlinear relationship. The ANOVA results show that the models satisfactorily predict the difficult-to-measure dimensions from the easily measured ones.