Dr. Stephen Lindsay, University of Victoria
Title: Replication in Psychological Science
Psychology journals have too often published Type I errors as real effects and too often greatly overestimated the sizes of real effects. As Interim Editor of Psychological Science, I am attempting to forge ahead on a path blazed by my predecessor Eric Eich: A multi-pronged effort to reduce the rate of such errors without sacrificing other scientific values (e.g., relevance, interestingness, importance). I highlight four issues that I believe particularly important: (a) the troubling trio of low statistical power, a surprising result, and a p value only slightly less than .05; (b) p hacking; (c) the noisiness of correlations; and (d) misinterpretation of non-significant effects. I encourage pre-registration, fine-grained graphing, power analysis, confidence intervals, effect sizes, replication, sharing of materials and data, and meta-analysis. My talk will allow for lots of back-and-forth and discussion about publishing in psychological science (including in Psychological Science).
