"On teaching statistical inference: What do p values (not) mean?"
with Bruce Blaine, St. John Fisher College
Hosted by: Leigh Slauson, Capital University
The misunderstanding and misuse of null hypothesis significance testing (NHST) for statistical inference is well documented, particularly among behavioral science students and researchers (Oakes, 1986). Central to this problem is the misinterpretation of p levels generated by significance tests. Two common errors made are: interpreting p as an effect size estimate, and interpreting p as the probability of the null hypothesis being true or false. Each misinterpretation can be exposed and examined with classroom data analytic exercises. The first involves showing students that increasing sample size reduces p levels but does not change effect size estimates. The second involves showing that a meta-analysis of "nonsignificant" studies produces a significant meta effect. These exercises serve a larger pedagogical goal of promoting estimation, as opposed to or along with, NHST methods for statistical inference.
Having trouble viewing? Try: Download MPEG-4 Podcast (.mp4)