Module 07 Summary

What this module established

The best effect size is usually the one that needs no explaining: the difference in units the reader already uses. A 15.1 ms saving, interval 4.5 to 25.7, tells a decision-maker more than any standardised measure.

Carry forward

• Cohen's d responds to the spread as much as to the gap. The same measure gave d = 0.28 and d = 0.74 depending on how variable the population was.
• Odds ratios and risk ratios agree only when the outcome is rare. With a common outcome a risk ratio of 1.5 came with an odds ratio of 2.25.
• Two studies can both reject the null and only one tell you the effect size — the small one's interval was 5.2 times wider.
• With 50,000 per group a 0.348-point difference was significant at p = 0.0002 and the entire interval sat below the 2-point bar. The study ruled a meaningful effect out.

Before moving on

Move on when you have stated the smallest effect worth acting on and compared the whole interval to it.

Mark complete