# Learning strategies: a synthetic comparison

This fictional randomized parallel-group study compared final knowledge-test
scores after active practice or rereading, with the same assumed learning
material and study time. All 80 synthetic participants contributed one score;
there were no missing values or exclusions. Active practice had a higher mean
score (n = 40, M = 67.47, SD = 10.19) than rereading (n = 40, M = 61.83,
SD = 11.30).

The planned two-sided Welch t-test estimated a **practice-minus-rereading
difference of 5.65 score points**, 95% CI [0.86, 10.44], t(77.19) = 2.35,
p = .021. Cohen's d was 0.53, calculated as the mean difference divided by
the pooled sample SD. The confidence interval above describes the
unstandardized mean difference; no interval for d was calculated.

The simulated design supplies the independence assumption. Distribution
summaries and the [boxplot](.nlss/runs/20260918T081434-30bad27d05/plots/figure-001-box-condition-score-condition.png)
show substantial within-group variation and overlap. Neither group's values
were flagged by the descriptive 1.5-IQR rule. The Shapiro–Wilk screens did not
detect departures from normality; this does not establish normality or
independence. Welch was selected in advance, without a preliminary-test rule.

Under the analysis assumptions, the estimated difference is positive but
imprecise: its interval spans roughly one to ten score points. No threshold
for educational importance was specified, so neither the p-value nor d
establishes practical value. A group-level difference also does not imply that
every learner benefits. Most importantly, these scores were generated from
assumed parameters; they are **no empirical evidence that active practice
outperforms rereading in real learners**.

Source context: [study sheet](study-sheet.md). Calculated evidence:
[descriptives](.nlss/runs/20260918T081414-bf45cc1f66/result.json),
[diagnostics](.nlss/runs/20260918T081421-18379fd17a0/result.json),
[Welch result](.nlss/runs/20260918T081427-247130f26b3/result.json),
[saved analysis request](.nlss/runs/20260918T081427-247130f26b3/request.json),
and [automatic protocol](report_canonical.md). NLSS 2.0.1 / R 4.5.2.
Reported means follow the protocol's two-decimal display; calculations use
unrounded values. Teaching material 1.0. © 2026 Mike Hammes / NLSS,
[CC BY 4.0](https://creativecommons.org/licenses/by/4.0/).
