Which intervention caused the observed improvement?
Create a factorial comparison with identical cases across configurations. Report interactions and uncertainty across cases.
Keep each case paired. Compute retrieval and review effects plus interaction = both − retrieval − review + baseline.
Cut these cards apart. Use the original example first, then let a partner supply a changed case. Keep sealed/final cards with the teacher until the learner commits a rule.
Case: A
Baseline: 0.4
Retrieval only: 0.6
Review only: 0.5
Both: 0.9
My prediction and reason:
Case: B
Baseline: 0.5
Retrieval only: 0.7
Review only: 0.6
Both: 0.75
My prediction and reason:
Case: C
Baseline: 0.2
Retrieval only: 0.3
Review only: 0.4
Both: 0.7
My prediction and reason:
Case: D
Baseline: 0.7
Retrieval only: 0.75
Review only: 0.8
Both: 0.82
My prediction and reason:
Case variation is not automatically independent sampling. Identify the population before interpreting uncertainty.