There are other worlds than these: why sensory and consumer science should embrace Bayesian inference

Sensory and consumer science aims to understand, describe, and predict people's experiences (sensory, emotional, hedonic , etc.) of stimuli or products through measurement and analysis of their reactions thereto. How well these analyses address our research questions is, however, rarely interrogated. We instead routinely rely on the traditional practice and associated inferential guidelines of frequentist null hypothesis statistical testing (NHST) and p -values. Here we propose a renewed discussion on how data analysis is handled within sensory and consumer science. Specifically, we posit that Bayesian inference is a valuable addition to our statistical toolbox. While NHST evaluates statistical hypotheses by asking “how likely are the data given the hypothesized (null) value?”, the Bayesian statistical framework flips this to “what is the relative compatibility of different values given the data and our pre-existing beliefs?” thereby directly addressing the questions we are often seeking answers to. In this article we: discuss common concerns with NHST, contrast the underpinnings of NHST and Bayesian inference, describe how Bayesian inference is conducted, directly compare the results of previously published data analysed using both approaches, and exemplify using Bayesian inference for two common cases in sensory and consumer science (determining discriminability and intensity equivalence). We also address common practical and analytical concerns with Bayesian statistics. Deeper appreciation for how statistical and scientific inference are related as well as increased awareness of and competence in different statistical approaches would enrich our research. Given the advantages the approach offers, this effort should include increased use of Bayesian inference.