Disrupting the Null Ritual via MOOCs: Can Dynamic Instructional Support Reduce Statistical Misconceptions?

• In Event: Moscone Center North Roundtable Session 10
  In Roundtable Session: Measurement and Aggregation in Online Learning

Abstract

Current distrust in research is in part motivated by the misuse of statistical methods, markedly in the case of null hypothesis significance testing (NHST). This systematic pattern of documented error has been often attributed to the formulaic approach to NHST (i.e. “the null ritual”), highlighting the need to replace mechanical rote instruction with dynamic and explicit training tools to reduce widespread statistical misconceptions. The current work investigates the effect of active instructional scaffolding to pinpoint and clarify common p-value misconceptions across an 8-week MOOC. Individuals receiving additional training show significant improvements in learning, maintained until week 8. Findings emphasize the real potential that dynamic training techniques, such as online learning platforms, may have on the effective improvement of statistical misunderstandings.

Authors

• Arianne Herrera-Bennett, Ludwig-Maximilians-Universität München

• Daniel Lakens, TU Eindhoven

• Moritz Heene, Ludwig-Maximilians-Universität München

• Stefan Ufer, University of Munich