ABSTRACT
This chapter presents an uncommon cross-section through the philosophy of statistics, namely one through the concept of modeling. The following text contains three parts that entertain three different—though related—perspectives on statistical modeling. The first part is devoted to the classical standpoint and the origins of the concept of a statistical model. A model mediates between mathematics, data, judgment, and economy of computation. The philosophical significance of this mediating role elucidates a controversy about modeling between the main proponents of the classical camp (Fisher, Neyman and Pearson). Part two discusses the counter-movement of “Exploratory Data Analysis” (EDA) led by John W. Tukey in the 1960s and 1970s who pleaded to abandon models and let the data speak for themselves. The section closes with a brief discussion of recent data-driven objections to modeling. Part three turns to the career of Bayesian modeling in statistical practice. It experienced a remarkable upswing in the 1990s, but at the same time, this success challenges Bayesian epistemology. The section closes with a brief look at recent accounts of practicing statisticians (of varying camps) who have (re)discovered the notion of modeling as a new focus and as a common ground.
