It is widely believed that robustness (of inferences, measurements, models, phenomena and relationships discovered in empirical investigation etc.) is a Good Thing. However, there are many different notions of robustness. These often differ both in their normative credentials and in the conditions that warrant their deployment. Failure to distinguish among these notions can result in the uncritical transfer of considerations which support one notion to contexts in which another notion is being deployed. This paper surveys several different notions of robustness and tries to identify why (and in what circumstances) each is valuable or appealing. I begin by discussing the notion of robustness addressed in Aldrich's paper (robustness as insensitivity of the results of inference to alternative specifications) and then discuss how this relates to robustness of derivations, robustness of measurement results, and robustness as a mark of casual as opposed to (merely) correlational relationships.
Stored only when both papers are in the corpus — 193 citations exist upstream.
Imported from OpenAlex (CC0) — nothing here is generated.