Commentary: Exploratory data analysis

Despite the importance of exploratory data analysis (EDA) in statistics and science, few people have worked on its philosophical foundations. In psychology, the present author (Haig, 2012; Behrens et al., 2013) have commented on philosophical aspects of EDA. They hold contrasting views about the appropriateness of abductive reasoning as a core component of the philosophy of EDA. Behrens and his co-authors think that abduction provides the “core logic” of EDA. I disagree. In this commentary, I say why I think their position is mistaken, and that their charge that mine is “a particularly disturbing” view of EDA is unfounded. Abduction as a form of inference is not well-known in academic circles. Broadly speaking, abduction is concerned with the generation and evaluation of explanatory hypotheses. In this sense, it contrasts with the more familiar ideas of inductive and deductive inference. Behrens et al. begin by taking their cue from Charles Peirce, and state that abduction is the form of inference involved in generating new ideas or hypotheses. However, surprisingly, Behrens et al. then elect to follow Josephson and Josephson (1994), and characterize abductive inference with the following pattern of reasoning (p. 39):