AtlasSession 4L10: Sample size calculationsp.13–14, 17–18
Power calculations
- 8 earlier topics
- CIs and p-values
- Beta error and power
- Why sample size matters
- Continuous outcomes
- Power calculations
Fig. 1
Bilayer dermal substitute vs split-skin graft alone: difference in mean patient-rated scar score (0–100, points)
Fig. 1. Two overlapping humps: the results you would get if there is no real difference, and if there is a real difference of 5.00 points. With 63 patients per group, about 80 in 100 studies would spot the real difference and about 20 in 100 would miss it. Spotting it about 80 in 100 needs 63 patients per group. Showing the plan.
- Showing
- the plan
- Studies that spot it
- about 80 in 100
- Studies that miss it
- about 20 in 100
- patients per group
- 63
- Needed to spot it about 80 in 100
- 63 patients per group
Two-Curve Test Visualiser · First look
Level 1 · First look · Lecture 10 · p.13–14, 17–18
Explain the difference between planning a study's size and checking afterwards whether it could still detect a real effect.
Distinguish a sample size calculation (design stage, expected inputs) from a power calculation (after observation, actual inputs) and judge whether β is still at or below the desired level.
Plan vs reality (the step you are on)
The trial planned 63 patients per arm, but only 40 per arm enrolled.
The plan: 63 patients per group, so that about 80 in 100 trials like this would spot a real benefit. Only 40 per group enrolled.