Membership Function Design for Multifactorial Multivariate Data Characterizing and Coding in Human Component System Studies

Whatever is performed using either an experimental or observational design, human component system studies involve several factors (seen as system inputs, with either quantitative or qualitative scales) and variables (outputs, with either quantitative or qualitative scales, objective or subjective aspects and time or nontime variables), where the study's main objectives are to infer both input/output and output/output relationships. Within this context, this paper explains the key role played by the membership function design (MFD) for an exploratory multifactor multivariate statistical analysis. First, several MFD options are compared using simulated examples, the method used to reach the study objectives being the correspondence analysis (CA). The CA benefits versus principal component analysis are shown. Then, three main taxonomic dimensions are considered for MFD: 1) mathematical/statistical/physical criteria; 2) monovariate/multivariate membership functions; and 3) unadapted/adapted approaches. Some MFD suggestions are then considered with actual studies such as eye movement in advertising, hand movement in tracking, hand position coordination in steering wheel turning, and subjective data in data entry task workplace assessment. The discussion section weighs the pros and the cons of using space windowing to perform a preliminary analysis of a large multifactor and multivariate database and provides some advice for MFD.