Most empirical studies that deal with activity analysis develop models of daily activity patterns in which the model is assumed to represent all of the days of the workweek. An alternative approach, in which activity pattern models are developed separately for each day of the workweek, is presented. The underlying assumption is that the appropriate basic time unit for analyzing some activity patterns is the week not the day. By applying this approach, a better representation of the behavior of individuals and improved models of activity patterns can be achieved. This hypothesis is tested by developingindependent models of daily activity patterns for each day of the workweek and comparing them among themselves and with an average-day model of shopping behavior. The results vary systematically during the week and thus encourage the development of day-of-the-week modelsfor analyzing activity and travel patterns. Furthermore, predictions based on average-day models were found to be biased when used in analyzing a specific day of the week.(A).
Abstract