Phase 1: Prepare
Step 1 Warm-up recall
List the four effect measures you have met: d, r, an odds ratio and a relative risk. Which one did Cohen and colleagues (2016) report?
Private: nothing you type here is sent or stored.
Step 2 Learning goal
By the end of this lesson you can:
- say what each effect measure describes (Foundation);
- avoid comparing numbers on different scales (Core);
- write a cross-module evidence summary and prepare week-4 data for the report (Core).
Phase 2: Teach
Step 3 A familiar example
Comparing a temperature in Celsius with one in Fahrenheit by looking at the numbers misleads. Effect measures also come in different units.
Step 4 Clear explanation
Each measure answers a different question:
| Measure | What it describes | Example in this course | No-effect value |
|---|---|---|---|
| Cohen’s d | Difference between group means in standard-deviation units | d = 0.28, kindness (Curry and colleagues, 2018) | 0 |
| Correlation r | Strength of a linear association | r = .29, positive activities (Sin and Lyubomirsky, 2009) | 0 |
| Odds ratio | Ratio of odds in two groups | OR 1.50, social relationships and survival (Holt-Lunstad and colleagues, 2010) | 1 |
| Relative risk | Ratio of risks (probabilities) in two groups | RR 0.83, purpose and mortality (Cohen and colleagues, 2016) | 1 |
Rule. Numbers on different scales cannot be compared directly. An odds ratio of 1.50 is not “bigger” than a d of 0.28 in any simple sense. To compare, convert to a common scale, such as a chance of superiority, or compare the verbal descriptions and the quality of the evidence.
Conversions you have used. d = 2r ÷ √(1 − r²). Percentage change in odds = (OR − 1) × 100. Percentage change in risk = (1 − RR) × 100. Chance of superiority = area below d ÷ √2.
Week 4 reminder. This is your last week of data. Rate at the same time each day, log each dose, and write a daily note on anything unusual. In Module 8 you will calculate adherence, the weeks 3 and 4 mean, the change from baseline and apply your decision rule.
Step 5 Worked example (4 marks)
A summary table lists “d = 0.28” for kindness and “OR = 1.50” for social ties, and a reader concludes ties matter more.
- The two are on different scales and cannot be compared by size. (1 mark)
- The outcomes differ: well-being in one, survival in the other. (1 mark)
- The designs differ: pooled trials versus pooled observational studies. (1 mark)
- Fair comparison: describe each in words, state the design and grade the evidence separately. (1 mark)
Phase 3: Practise
Step 6 Learner practice (instant-marked)
3. Which measure has a no-effect value of 1?
Ratios are 1 when the groups are equal; d and r are 0.
4. Why is it misleading to say OR 1.50 is a bigger effect than d 0.28?
Different scales cannot be compared by size.
Phase 4: Close
Step 7 Recap
- d and r have a no-effect value of 0; odds ratios and relative risks have 1.
- Do not compare numbers on different scales directly.
- Week 4: rate, log and note; change nothing.
Step 8 Exit ticket (ungraded)
Write one sentence comparing two findings from this course without comparing their raw numbers.
Private: nothing you type here is sent or stored.
Step 9 Go further (Extension)
Make a one-page table of the effects you have met, adding design, size of sample and your grade.
Common mistakes
| Mistake | Why it fails |
|---|---|
| Ranking effects by raw numbers | The scales differ. |
| Mixing up the no-effect value | It is 0 for d and r, 1 for ratios. |
| Ignoring design when comparing | Trials and observational studies differ in what they can show. |
Limits. This lesson summarises conversions used in the course; the formulas assume simple cases. Figures quoted are from earlier lessons.