Understanding group comfort through directional statistics

Abstract

This paper has the dual aims of introducing and using directional statistics to investigate the comfort levels of pairs of people approached by a robot from different directions. Data from pairs seated in three maximally-different seating configurations are analysed. These data are in the form of circular distributions of ranked comfort levels. Statistical tests for uniformity of circular distributions and for determining if differences exist between pairs of circular distributions are introduced and used to analyse the directional properties of the data. It is shown that directional statistics can be used to compare comfort level ranks that capture all tested robot approach directions; something that cannot be achieved with linear statistics.