A Factorial Model of Aggregate Spatio‐Temporal Behavior: Application to the Diurnal Cycle

The cross‐sectional nature of much social data, coupled with the static view provided by maps and current spatial data handling software, have produced a tradition of research on urban spatial structure that is largely two‐dimensional and derived from residential locations. The paper presents an analysis of a space‐time diary data set collected in Halifax, Nova Scotia. A series of transformations are used to convert the individual diary records to a three‐mode matrix of intensities, which is then analyzed using the PARAFAC three‐mode factor model. Home/work is found to be the strongest organizing dimension of the urban space‐time, followed by entertainment, shopping, and education / work. We show how these dimensions appear to varying degrees in different locations, time periods, and human activities. The paper argues for a dynamic view of urban spatial structure in which only the physical facilities remain static. 1993 The Ohio State University