Theory construction and data analysis. (February 2017)

• Record Type: Journal Article
• Authors: Tryon, Warren W.; Patelis, Thanos; Chajewski, Michael; Lewis, Charles
• Is Part Of: Theory & psychology. Volume 27:Number 1(2017:Feb.)
• Page Start: 126
• Page End: 134
• Publication Date: 2017-02
• Subjects: falsifiability -- null hypothesis statistical testing -- publication bias -- replicability -- theory construction -- web of causation; Psychology -- Periodicals; 150.1
• DOI: 10.1177/0959354316684043
• Languages: English
• ISSNs: 0959-3543

Abstract

Data analysis should be neutral relative to theory construction in order to be unbiased. Data analysis should not strongly favor one form of theory construction over others. Traditional approaches to theory construction prioritize simplicity of explanation based on parsimony using a few prominent statistically significant variables. Alternative web of causation explanations prioritize comprehensive explanations based on complexity using many small effects. This article presents argument and empirical evidence that contemporary data analytic methods are problematic for all theory construction approaches. They are especially biased against web of causation approaches to theory construction. Mathematical arguments and new empirical evidence that supports web of causation explanations are presented.