Effect sizes

Question 1

effect size

Answer 1

• express the difference as the number of standard deviations
• = Cohen’s d

Question 2

Cohen’s d

Answer 2

= (mean intervention - mean control) /SD

Question 3

which standard deviation?

Answer 3

• pooled SD from intervention and control
• but also weigh for sample size in each group
• use computer program to calculate

Question 4

interpretation of cohens d, numbers

Answer 4

0 = no difference
0.2 = small
0.5 = moderate
0.8 = large 
based on rule of thumb

Question 5

how do you interpret a Cohens d of 0?

Answer 5

there is no effect of the treatment (it is 0 SD better than the other group)

Question 6

interpretation of Cohens d

Answer 6

1. magnitude of effect
2. independent of specific questionnaire
3. you can compare effects of different RCTs using different questionnaires

Question 7

Why does the effect size not tell us anything about the significant?

Answer 7

• significant depends on sample size (hence power analysis)

Question 8

what is a 95% CI?

Answer 8

you can be 95% sure that the true effect is within this interval

Question 9

example of CI

Answer 9

e. g. Cohens d = 0.50 (95% CI 0.30 to 0.70)
- the effect in your RCT (point estimate) = 0.50
- you can be 95% sure that the real effect lies between 0.30 and 0.70)

Question 10

what can the CI tell us about significance?

Answer 10

CI cannot include 0, then it is significant (with 95% probability you can say that your effect lies within this interval)

Question 11

how to interpret an effect size plus CI

Answer 11

1. the intervention group scores … SD more than the control group (point estimate)
2. the true effect lies between … and … SD
3. the result is (not) statistically significant (does (not) contain 0)
   - the smaller the CI, the more certain you are about the effect, you had a larger sample

Question 12

what does a negative effect size mean?

Answer 12

• the control group might have scored better

- or lower scores are better