We’ve changed the way we discuss statistical inference, p-values and confidence intervals. The debate over the arbitrary 5% cutoff that is considered “significant” has been rumbling on for many years, but it was given fresh impetus in Nature last year with an editorial titled, “It\u2019s time to talk about ditching statistical significance”. This was in response to an article by Valentin Amrhein, Sander Greenland, and Blake McShane (backed by more than 800 signatories) that criticises the pervasive reliance on using p-values in this way (see “Scientists rise up against statistical significance”).\n

Section 2.3, particularly the “Find out more” box, is an example of how we discuss p-values now. In short: instead of asking students, “Is the difference in means statistically significant?”, we encourage students to ask themselves, “What do the p-values tell us about the difference in means?”\n

If the commitment not to put little stars next to empirical results fills you with dread, watch this explanation by Margaret Stevens, of Core Economics website’s steering group, about why we made the change. She gave this presentation during the Doing Economics session of Core Economics website’s 2019 teaching and learning workshop. The discussion of p-values and significance starts at 6:38, but if you are interested in using Doing Economics in your course, the whole session is worth a watch.\n