Some Post-Presentation Thoughts

· Don’t forget about confidence intervals.  They are especially valuable after you do a test and find a statistically significant result.  Then, instead of just saying “we have evidence men and women differ,” you can say “on average, the male population has scores this much higher than the female population” – being very clear what we are hoping to capture inside the confidence interval.

· Don’t forget about your numerical and graphical summaries.  After you come to a conclusion from your test of significant, refer back to your descriptive statistics to see if the results are consistent, e.g., with have a large p-value and as we say there is a lot of overlap between these sample distributions. (And often you are still not including enough discussion of the descriptive statistics the first time you introduce them either.)

· Several of you have gotten into the habit of saying “the p-value is small so we reject and the test statistic is large, providing even more evidence.” I don’t really see this as “more” evidence.  It is nice to check the consistency between these results, but it is really the same piece of information.  The p-value can’t be small without a large test statistic.  I insist that you report the test statistic to help me know where your p-value came from. Once we got into F statistics and Chi-square statistics, we lost our intuition for how large is large (no longer able to interpret these as “number of standard errors between observed and hypothesized”) which also depends on the sample size, so we need the p-value to decide if we have obtained a surprising value or not.