On de-bunking “Fake News” in the post-truth era
How to reduce statistical error in research
Bent Flyvbjerg (University of Oxford)
Atif Ansar (University of Oxford)
Alexander Budzier (University of Oxford)
Søren Buhl (Aalborg University)
Chantal Cantarelli (University of Sheffield)
Massimo Garbuio (University of Sydney)
Carsten Glenting (Viegand Maagøe A/S)
Mette Skamris Holm (Aalborg Municipality)
Dan Lovallo (University of Sydney)
Eric Molin (TU Delft - Technology, Policy and Management)
Arne Rønnest (The National Center for Coastal Fishing and Angling)
Allison Stewart (Infrastructure Victoria, University of Oxford)
Bert van Wee (TU Delft - Technology, Policy and Management)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
The authors note with alarm that statistical noise caused by statistical incompetence is beginning to creep into research on cost overrun in public investment projects, contaminating research with work that does not meet basic standards of validity and reliability. The paper gives examples of such work and proposes three heuristics to root out the problem. First, researchers who are not statisticians, or do not have a strong background in statistics, should abstain from doing statistical analysis, and instead rely on more experienced colleagues, preferably professional statisticians. Second, journal referees should clearly state their level of statistical proficiency to journal editors, so these can set the right referee team. Finally, journal editors should make sure that at least one referee is capable of reviewing the statistical and methodological aspects of a paper. The work under review would have benefitted from observing these simple heuristics, as would any work based on statistical analysis.