A Bayesian Nonparametric Test of Significance-Chasing Biases
19 Pages Posted: 18 Aug 2017 Last revised: 22 Sep 2017
Date Written: August 21, 2017
There is growing concern that much of the published research literature is distorted by the pursuit of statistically significant results. In a seminal article, Ioannidis and Trikalinos (2007, Clinical Trials) proposed an omnibus (I&T) test for significance chasing (SC) biases. This test compares the observed number of studies that report statistically-significant results, against their expected number based on study power, assuming a common effect size across studies. The current article extends this approach by developing a Bayesian nonparametric (BNP) meta-regression model and test of SC bias, which can diagnose bias at the individual study level. This new BNP test is based on a flexible model of the predictive distribution of study power, conditionally on study-level covariates which account for study diversity, including diversity due to heterogeneous effect sizes across studies. A test of SC bias proceeds by comparing each study's significant outcome report indicator against its estimated posterior predictive distribution of study power, conditionally on the study's covariates. The BNP model and SC bias test are illustrated through the analyses of three meta-analytic data sets, and through a simulation study. Software code for the BNP model and test, and the data sets, are provided as Supporting Information (by request of the first author).
Keywords: Bayesian Nonparametrics, Meta-Research, Significance Chasing Bias, P-hacking, Publication Bias, Bernstein Polynomials
Suggested Citation: Suggested Citation