Retail marketing geography has traditionally employed static gravity models for location analytics based on probabilistic locational consumer demand. However, such retail trade area models provide little insight into the dynamic space-time hierarchical diffusionary processes that aggregate to an eventual market structure equilibrium (Mason et. al., 1994), which gravity models attempt to predict for retail trade areas. In addition, most attempts to display the aggregating dynamic space-time hierarchical diffusionary processes of space, time and attributes of interest, in a geographical information system (GIS), produce visualizations that are overly complex and typically displayed utilizing unfamiliar paradigms. Further, these attempts fail to take into account the extensive body of literature in psychology and brain science that stress the importance of perceptual elements and design in achieving optimum visualization comprehension. In other words, simplicity (three-way factor analysis) and visual familiarity (cognitive fit theory (Vessey, 2006), mere-exposure effect in psychology (Dajonc, 1968). This will provide faster perception and better visuospatial and temporal understanding of objects and trends. In this study we incorporate these elements in our visualization object that we refer to as “Avatar”. A Huff inspired, Bayesian framework of inference for spatial allocation and hypothesis testing allows the Avatar object to display the spatial allocation of the Bass model’s innovators and imitators for sales forecasts of new product diffusion (e.g. a mathematical version of Everett Roger’s adoption concept), thus enabling and supporting faster and improved visuospatial understanding of very large data repositories of unbounded and/or “countably infinite” sized geo-big-data (referred to throughout the rest of this paper as GBD). We then introduce the three steps necessary to create an Avatar object (i.e. a 3-D semaphoric, space-time diffusion visualization object). The Avatar object is designed specifically to visualize determinant attributes (e.g. demographics) for the Bass, Bayes, Berry and Huff integrated ensemble model forming part of an ancillary paper to this study. In this way we display the timed hierarchical diffusion of new innovative products throughout store trade areas and across the ensuing and evolving store networks. In addition, by calculating Bayesian conjugate priors and posterior spatial allocation probabilities for the “smallest units of human settlement” (Christaller, 1966) or in our case statistical demographic units (i.e. Census Blocks), we establish customer (innovator and imitator) spatial distributions for the Bass temporal-only model for the case of the aggregating store level trade area (SLTA) scenario. Our approach is empirically supported by five years of new product diffusion geocoded panel data from the Southern California market. We conclude that our cognitive fit theory validated Avatar space-time diffusion visualization strengthens “location analytics” and “location intelligence” and provides a simple and familiar tool for displaying GBD across a growing domain of varying applications and end-user knowledge and needs.
Franklin, Christopher, "Space-Time Diffusion Visualization using Bayesian Inference" (2015). 2015 Proceedings. 4.