The sort of realism supported by statistical principles and experimental design is a thin one. Populations or their members have a real value independent of attempts to measure them, and independent of observations concerning them. As a result, the realist can invoke these unobserved, real values in their explanations of observed behavior or correlations. That there could be a general argument for realism required to account for the reliability of specific statistical principles is a novel contention. Although it is a familiar idea that the theoretical considerations raised in good experimental design require a realist interpretation, no such presumption in favor of realism has accompanied the treatment of general statistical principles. Robust realism, by contrast, holds that the approximate truth of mature scientific theories is the best explanation for their success. This strong version of realism has only mature sciences as its subject matter, and the evidence for it in the social and behavioral sciences is uneven.
Keywords: experimental design, statistical principles, realism, observation, theoretical, mature sciences, robust realism, minimal realism