This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. For attribution, the original author(s), title, publication source (PeerJ Preprints) and either DOI or URL of the article must be cited.
In a recent paper, Beall, Hofer, and Schaller (2016) use observational time series data to test the hypothesis that the 2014 Ebola outbreak influenced the 2014 U.S. Federal Elections. They find substantial associations between online search volume for Ebola and people’s tendency to vote Republican, an effect observed primarily in states with norms favoring Republican candidates. However, the analyses do not deal with the well-known problem of temporal autocorrelation in time series. We show that all variables analyzed exhibit extremely high levels of temporal autocorrelation (i.e. similarity in data-point values across time). After appropriately removing first-order autocorrelation, the observed relationships are attenuated and non-significant. This suggests that either no real associations exist, or that existing data are insufficiently powered to test the proposed hypotheses. We conclude by highlighting other pitfalls of observational data analysis, and draw attention to analytical strategies developed in related disciplines for avoiding these errors.
Subject Areas updated to include "Statistics" & "Psychiatry and Psychology".