Papers and books claiming to provide scientists with statistical methods for “causal inference” are increasingly seen in the ecological literature. Interest in drawing causal inferences from data is central to all scientists; thus, it is important for practitioners to understand the assumptions that lie beneath recommendations of what to do and what not to do that are being offered. In this paper I will focus on (a) the foundational premise of causal statistics, (b) the dilemma it creates for practitioners, (c) some of the work-arounds being purported as solutions, (d) questionable claims being made that illustrate the ultimate vulnerability of those solutions, and (e) a potential remedy that brings competing approaches under a common framework and provides for a clearer understanding of the merits of different approaches.
Results/Conclusions
The foundational precept of causal statistics is that it is impossible to draw causal inferences from data except under the rarified conditions of an ideal randomized experiment. The reasoning is that it is impossible to guarantee causal identification (complete absence of confounding or bias) under any other circumstances. In the last 20 years, a fusion of techniques has morphed into what some methodologists see as a limited reprieve from the ban on causal inferences. This presumed reprieve is seen to only apply under an extremely narrow set of conditions. Following the newly-offered prescription leads to the acceptance of impossible restrictions on science and obviously incorrect classifications of methodological approaches. It also includes a “poison pill” assumption that is fatal to all approaches to causal inference from data. Resolving this situation requires a critical examination of the foundational precept of causal statistics and a broader conceptual framework, along with new definitions and distinctions. Implementing these brings the full suite of quantitative traditions for inferring causal effects under a common framework that supports a comprehensive approach for the advancement of science.