On the dialogue between ethnographic field work and statistical modelling

Ethnographic and statistical methods have very different epistemological underpinnings; ethnographic research is interpretive, while statistical research is generalizing. Our case study combines these two contrasting approaches and brings them into an eye-level conversation with each other. We conceptualized a Bayesian statistical model for Rainwater Harvesting Mode in rural South Africa, based on hypothetical relations derived from ethnographic field observations. The model pointed to spurious relations and that new hypotheses from fieldwork helped explain. Mixing the two methods means using one to critically reflect on and challenge the other, lending robustness to the research process and the results in a form of triangulation. Iterating ethnographic field work and statistical modelling is useful for learning about particular places. Importantly, we do not see statistical modelling as the end point that ethnographies may provide hypotheses for, but as a recurring step of quantification that generates valuable questions for subsequent ethnographic field work.