Accounting Research as Bayesian Inference to the Best Explanation

Abstract The problems with p -values have been extensively discussed recently, but there is little work about the broader aspects of scientific inference of which p -values are but one part. This article explains how scientific inference can be characterized as Bayesian inference to the best explanation, which involves developing and assessing theories based on their fit with background facts and their ability to explain the observed data better than competing theories can. These factors translate into prior odds and Bayes factor respectively, which determine posterior odds under Bayesian inference. I provide examples from accounting research to illustrate how attention to these points makes for better research designs and stronger justification for inferences.