Uhl, Alfred (2019): Problems when interpreting correlation causally. 30th Annual Conference of the European Society for Social Drug Research, 26. September 2019, Riga.

Full text not available from this repository.

Abstract

OBJECTIVES: To demonstrate unequivocally that the popular intuitive interpretation of correlations as causal relationships in policy discourses is commonly inadequate? *** METHODOLOGY: To search for causal interpretations of correlations in the alcohol policy discourse, which are usually presented and accepted unquestioned and to show as clearly and unambiguously as possible that the conclusions are not only problematic but definitely unsustainable with convincing examples. *** SIGNIFICANT RESULTS: The tendency to treat correlations of time series as solid prove for causal relationships is very strong when policy interventions are to be evaluated. This is particularly true in the so called “evidence-based” alcohol policy discourse. The presentation provides examples, where causal conclusions seem very convincing and confronts with similar examples where it is easy to grasp these conclusions are not justified. *** CONCLUSIONS: When it comes to causal statements based on observation, competent, critical questioning of these conclusions is indispensable. Paradoxically the desired "pure observation" can only be achieved by deconstructing the immediate perception with an intellectual (analytical) endeavour, i.e. doing an "interpretation" based on knowledge (Kriz et al. 1990). Causal interpretations derived from empirical data must not be interpreted as false logical conclusions but these must be explained as "Gestalt" phenomena, similar to optical illusions, which can hardly be avoided deliberately by recipients. As Michotte (1982) stated: “We experience causality as direct as we experience colours”. *** It must be added here that identifying illogical inference based on empirical data does not necessarily mean that the intuitive conclusions are factually wrong. The insight that specific causal inference is wrong only implies that the data under consideration neither prove nor disprove the conjectures these data seem to support and not that these conjectures are wrong - just as a court acquittal for lack of evidence does not imply that the defendant is innocent.

Item Type: Conference or Workshop Item (Lecture)
Subjects: OEBIG > Kompetenzzentrum Sucht
Date Deposited: 23 Mar 2020 19:18
Last Modified: 23 Mar 2020 19:18
URI: https://jasmin.goeg.at/id/eprint/1258