Beyond P<0.05 : Scientific Inference in Accounting Research

Although the limitations of p-values are well known, researchers continue to (mis)use them for lack of awareness of the alternatives. This monograph comprehensively explains the process of scientific inference, of which the p-value is just one component. The inference process is broader, consisting of research topic choice, assessing prior odds of the tested theory, choosing research designs having high diagnostic ability, and interpreting the results. The process follows the Bayesian scheme of inference, although its elements are not quantified numerically. The best researchers have always followed this process. Aimed at Ph.D. students and rising scholars, the monograph presents the issues in a non-technical language, with plenty of examples from accounting and other disciplines. It helps scholars to choose and motivate their research topic, identify better research designs, and communicate their findings. Technical topics covered include well-known ones such as confidence intervals, as well as less-well-known ones such as Bayesianized p-values, false discovery rates, and hierarchical Bayesian models. They are presented in a non-technical manner, with applications. All computer code used is available online. Qualitative topics covered include how to choose research topics and designs, and write the paper like a story. The monograph urges researchers to choose settings with a high signal-to-noise ratio, rather than get statistical power through large samples. The monograph’s call to return to basics is expected to make research more pleasurable as well as more insightful, and more relevant to practice