Two measurements on the same people, and the question of whether they moved. What the test answers, how the data has to be laid out, how it runs in Quanta on the Mac, and how to report it in APA 7.
A paired-samples t test compares two measurements taken from the same people. Pre-test and post-test. Before a workshop and after it. The same clinician rated on two occasions. Each row contributes a pair, and the test works on the difference within each pair rather than on the two group averages.
That is the whole reason to use it. Because each person serves as their own comparison, the differences between people fall out of the calculation. A paired design removes a source of noise that an independent-samples test has to absorb, which is why it detects a real change with far fewer participants.
The test asks whether the average difference across pairs is far enough from zero to be more than sampling variation. Use it only when the pairing is real. If the two sets of scores come from different people, the pairing is an illusion and the correct test is the independent-samples t test. If you have three or more repeated measurements on the same people, a paired t test on selected pairs will inflate your error rate, and repeated-measures ANOVA is the right tool.
Real pairing. Every row must be one person or one unit measured twice, in a consistent order. This is a property of your design, so no software can verify it. It is the assumption you own.
Normality of the differences, not the scores. This is the condition researchers most often get wrong. The paired t test makes no assumption about whether your pre-test or post-test scores are normally distributed. It assumes the differences are roughly normal. Two skewed distributions can produce perfectly well-behaved differences, and two normal-looking distributions can produce differences that are not. Check the differences.
Complete pairs. A pair with one measurement missing contributes nothing and is dropped. If dropout is heavy, the surviving pairs may no longer represent the sample you started with, which is a question about who your result describes rather than a question about the test.
Where the differences are badly non-normal or the outcome is ordinal, the Wilcoxon signed-rank test is the nonparametric counterpart and is included in Quanta.
Lay the data out in wide format. This is the opposite of what an independent-samples t test wants, and it is where most paired analyses go wrong before they start. One row per person, and the two measurements in two separate columns, for example Pre_Score and Post_Score. If your data is currently long, with one row per measurement and a time column, reshape before you run.
Bring in the data. Import from CSV, Excel, SPSS .sav, Stata .dta, or a ReliCheck survey package. Quanta reports the row count, the column count, and the missing-cell count as the file lands, which on a paired design is worth reading carefully. Missing cells are broken pairs.
Open Compare Means and choose Paired t-Test. Compare Means holds four analyses on one segmented control: t-Test, Paired t-Test, One-way ANOVA, and ANCOVA. Picking one changes what the Setup panel asks for.
Configure the variables in the Setup panel on the right. A paired test asks for the two measurement columns rather than an outcome and a grouping variable. Order matters for the sign of the result: whichever you nominate first determines whether an improvement reads as a positive or negative difference. Results appear in the center.
The result arrives on four tabs: Test result, Effect size, Diagnostics, and Reporting. Every tab carries the same bottom bar, so you can add to report, copy the table, or copy the APA text from wherever you are standing.
Test result gives the APA-shaped table, badged with the test that ran, and reports t with its degrees of freedom and p. Alongside the frequentist result, Quanta reports a JZS Bayes factor with the prior stated. On a paired design this earns its keep: an intervention study that fails to reach significance is a very different paper depending on whether the evidence favors no change or is simply too thin to say, and only the Bayes factor separates those two.
Effect size gives Cohen's d as a headline number with a magnitude word attached, rather than a bare figure you have to interpret from memory.
Diagnostics reports the assumption checks as sentences with verdicts, not as bare test statistics to be looked up.
Reporting writes the finding in APA style, ready to lift, with an explicit guardrail alongside it. A change over time is not by itself evidence that your intervention caused it, and the summary says so rather than leaving you to remember.
Every result carries a ReliCheck Intelligence card with one action: explain this result in plain language. It runs entirely on your Mac with Apple Intelligence. No upload, no API key, nothing to configure. The model only explains the result Quanta has already computed and validated. It never computes a statistic, and it never applies a decision on your behalf.
Copy APA text puts the result line on your clipboard. Add to report sends the table into a Quanta report, which exports with the APA table, its note, and a per-analysis caution beneath it.
Scores rose from pre-test (M = 64.20, SD = 8.90) to post-test (M = 71.05, SD = 8.14), a paired-samples t test showing a significant increase, t(58) = 5.94, p < .001, d = 0.77.
Report the number of pairs, not the number of measurements. A study of 59 people measured twice has n = 59 and 58 degrees of freedom, not 118. State how many pairs were dropped for incomplete data and why, because a reader cannot judge the result without knowing who is missing from it.
Where the result is null, report the Bayes factor with it. Not significant is an absence. Moderate evidence for no change is a finding.
Quanta's t tests, ANOVA, and ANCOVA are validated against R's t.test, aov, and car, matched exactly. The engine is pure Swift with no R or Python at runtime. Full record on the validation page.
Compare Means is part of every Quanta subscription, $9.99 a month or $79.99 a year after the 30-day trial. One price for everyone. See pricing.
The analysis, the report builder, and the on-device explanations all work with the network off. Your dataset is never uploaded.
Independent-samples t test for two separate groups · One-way ANOVA for three or more groups · Wilcoxon signed-rank when the differences are skewed · Friedman for three or more repeated measurements · the complete list on the analyses page.