Pre/post results
Did learners’ scores change between the pre-test and the post-test? One row per measure.
In short
Who went up
For each measure, the share of learners with both tests whose score went up, stayed the same or went down. Hover a bar for the count.
By measure
Only learners with both a pre-test and a post-test score for a measure are counted in that row, so each child is compared with themselves.
| Measure | Learners | Pre-test mean (SD) |
Post-test mean (SD) |
Change mean [95% CI] |
Up / same / down | Test | Result |
|---|---|---|---|---|---|---|---|
| Loading results... | |||||||
How these numbers are worked out
- Pairs. For each measure, every learner with both a pre-test and a post-test score is one pair. The change is post-test minus pre-test, so a positive change means the score went up.
- Normality. The Shapiro–Wilk test (Shapiro & Wilk, 1965; Royston’s 1995 algorithm, as in R, SPSS and SciPy) checks whether the changes look normally distributed.
- The test. If they do (Shapiro–Wilk p ≥ .05), the paired-samples t-test is reported, with Cohen’s d for paired data. If they do not, the Wilcoxon signed-rank test (Wilcoxon, 1945) is reported instead, with the matched-pairs rank-biserial correlation r (Kerby, 2014). The downloaded results include both tests for every measure.
- Significance. Two-tailed, at α = .05. Effect sizes are labelled by Cohen’s (1988) conventions: d of .2, .5 and .8 and r of .1, .3 and .5 are small, medium and large.
- Too few. A measure with fewer than three pairs is not tested.
The paired-scores download lists every pair with a learner code instead of a name, so the same tests can be re-run in Excel, SPSS, JASP or R.