“Can Bayes "Prove" the Null? What Should Replace "p < .05"?”
by Professor Michael Smithson
An attraction of Bayesian statistical methods for some researchers is a belief that, unlike the p-value in null-hypothesis tests, a large Bayes factor in favour of the null hypothesis and against its alternative is strong evidence for the null hypothesis. However, things aren't that simple, and it turns out that this belief can be misleading.
An investigation into the current views among statisticians regarding hypothesis-testing reveals that although everyone agrees that we should no longer use null-hypothesis significance testing (NHST), there are "significant" disagreements among them about what should replace it. This seminar provides a cautionary survey of alternatives to NHST.
Tuesday October 15, 4:00 – 5:00pm
Peter Baume Building 42A, Room 2.01
Followed by drinks and nibbles in the Research School of Psychology tea room, Building 39.