Share:
Configurational modelling involves simple but powerful methodologies that seamlessly integrate the design process and has high adherence by professionals. However, a traditionally intuitive approach, rather than a statistically informed one, occasionally compromises such models. As a consequence, the models often do not reach statistical significance and therefore are of limited efficacy. Were they to reach statistical significance, configurational models would increase their validity but also the potential for data sharing and therefore their economic feasibility. In this paper, I discuss the aspects of the process of creating configurational models that are crucial towards statistical validity. After introducing the methodological framework, I focus on the two stages for measuring traffic: finding a representative sample for each spatial unit and measuring sufficient units to form the model. I present new evidence of the expected variability of data that dismisses common but false assumptions that often lead to statistical insignificance. I demonstrate how the introduction of variance breaks down a model. I argue that reaching statistical significance can be achieved with the use of basic statistics, which are within reach of designers. Finally, I introduce Bootstrapping, an advanced but straightforward method to provide statistical significance in cases of a small sampling.
