Paired t-Test Studio

Updated Feb 2026 Paired means

Compare two related measurements with rich narratives, diagnostics, and visuals ready for stakeholder decks.

QUICK START: Choose Your Path

DATA SOURCE

Load a marketing use case:

Select a preset to auto-fill inputs for common measurement tasks (matched market uplift, pre/post surveys, creative benchmarks). Each preset references the raw CSV so you can inspect the required formatting.

Drag & Drop raw data file (.csv, .tsv, .txt, .xls, .xlsx)

Two numeric columns of paired measurements (before, after), or one column of differences (after minus before), with a header row. Up to 5,000 rows; blank rows are ignored.

No file uploaded yet.

INPUTS & SETTINGS

Type or paste your Before & After values directly (supports two-column paste from Excel/Google Sheets). Set how many rows you need (max 50) and enter values inline.

Row count (max 50)

Row

When you only have summary tables, supply the mean and standard deviation of the difference scores plus the sample size.

Mean difference SD of differences Sample size (pairs)

Analysis Settings

Significance level (α)

DECISION

⏳ Awaiting Data

Enter paired data or summary statistics to see if the mean difference is significantly different from zero.

t-statistic

—

p-value

—

α (threshold)

0.05

Mean Diff

—

Cohen's d_z

—

VISUAL OUTPUT

Provide data to summarize the mean difference and confidence interval.

💡 How to read the Mean Difference Fan Chart

Center marker (d̄): Shows the average within-pair change (After − Before) in your outcome's units.
Nested fan bands (50% / 80% / 95% CI): Darker inner bands show where the true mean change most likely sits; the widest outer band is your formal confidence interval at 1 − α.
Dashed "No change" (0) line: If the outer confidence band sits completely to the right (positive lift) or left (decline) of 0 without crossing it, the paired change is statistically significant at your chosen α.

💡 How to read the Distribution & Slopegraph views

Histogram of Differences: Plots each pair's change score (d = After − Before). Bars to the right of the red dashed 0 line represent units that increased; bars to the left decreased. The paired t-test assumes this distribution is roughly bell-shaped without extreme outliers when n is small.
Before → After Slopegraph: Connects each unit's starting value to its ending value (green = increase, red = decrease), overlaid with the bold navy group mean trajectory. Parallel lines indicate a uniform shift across units; steeply crossing lines reveal heterogeneous responses.

TEST RESULTS

💡 How to Interpret These Results

Reading the t-statistic and p-value:

t-statistic: Measures how many standard errors the mean difference is from zero. Larger |t| = stronger evidence of a real difference.
p < 0.05: Strong evidence that the mean difference is not zero. Reject H₀ (no change between conditions).
p ≥ 0.05: Insufficient evidence. Cannot conclude the before/after conditions differ significantly.
Very small p (< 0.001): Very strong evidence. The observed difference is highly unlikely due to chance.

Understanding the confidence interval:

The CI shows the plausible range for the true population mean difference
If the CI excludes zero, the result is statistically significant at that confidence level
Wider intervals = more uncertainty (often from smaller samples or higher variance)
The interval width matters for practical decisions—a significant but tiny effect may not be actionable

Effect size (Cohen's d_z):

d_z < 0.2: Small effect—difference may be statistically detectable but practically minor
d_z ≈ 0.5: Medium effect—noticeable difference that stakeholders can act on
d_z > 0.8: Large effect—substantial difference with clear business implications

Effect size helps contextualize significance: a large sample can detect tiny, meaningless differences as "significant."

Analysis Status

Enter data to see the paired t-test.

Paired t-Test

t(--)=--

p = --

95% CI: [--, --]

Effect sizes

Cohen's dz = --

Hedges' g = --

Mean difference = --

❓ What's this? (Cohen's d_z & Hedges' g)

Cohen's d_z (d̄ / s_d) expresses the average change in standard-deviation units of the within-pair differences: 0.20 = small, 0.50 = medium, 0.80+ = large / highly consistent shift.

Hedges' g is the small-sample bias-corrected version of Cohen's d_z. When n < 30, sample standard deviations slightly exaggerate d_z; Hedges' g trims that upward bias.

Mean difference (d̄) is the unstandardized average change (After − Before) in your original measurement units.

Overview

n = -- pairs

Mode: Paired columns

APA-Style Statistical Reporting

Summary will appear after analysis.

Managerial Interpretation

Business-facing copy will appear after analysis.

OVERVIEW & WHAT THIS TOOL SOLVES

When you measure the same customers, stores, or DMAs before and after a marketing intervention, running a standard two-sample t-test throws away the pairing and buries your treatment lift inside baseline differences between big and small units. Paired t-Test Studio computes each unit's within-pair change first, stripping out between-unit noise to test whether the average shift is significantly different from zero.

🎯 Questions It Answers

Did average spend, engagement, or store sales change significantly from Before to After within the same matched units?
How large is the within-unit shift relative to the variability of individual changes (Cohen's d_z)?
Are the pair-level differences clean enough for a parametric t-test, or do outliers call for a Wilcoxon signed-rank check?

📂 What Data You Need

Summary Mode: Mean of paired differences (d̄), SD of differences (s_d), and number of pairs (n)
Raw Paired Data: Two numeric columns (Before and After) measured on the exact same n subjects or stores
Difference Direction: Each difference is always After − Before, so a positive mean difference means the measure went up. In an uploaded file the first numeric column is read as Before. To flip the sign, swap the two columns.

💡 What You'll Discover

Paired t & Confidence Interval: Exact t-statistic, degrees of freedom (n − 1), p-value, and CI on the mean change
Standardized Effect Sizes: Both Cohen's d_z and small-sample bias-corrected Hedges' g_z
Pair-Level Visuals: Distribution of within-pair differences and before-vs-after concordance plots

🔍 Key Nuances & Common Traps to Watch For

Why Paired Beats Independent on Matched Data: When baseline store sales correlate strongly between periods, subtracting Before from After cancels out permanent store-size variance and boosts statistical power dramatically.
Normality Applies to the Differences, Not the Raw Columns: Even if store revenues are right-skewed across locations, the paired t-test only requires that the within-store differences (d_i) be approximately normal (or n ≥ 30).
Watch for Carryover & Seasonality: A significant before/after jump proves that outcomes changed over time, not necessarily that your campaign caused it—pair with a control group (Difference-in-Differences) when external macro trends are active.

📐 Under the Hood: Model Equations & Variable Definitions

1. Within-Pair Difference & Paired t-Statistic

\[ d_i = y_{\text{After},i} - y_{\text{Before},i}, \qquad \bar{d} = \frac{1}{n}\sum_{i=1}^{n} d_i, \qquad t = \frac{\bar{d} - \Delta_0}{s_d / \sqrt{n}} \]

💬 Read it in plain English: We first subtract each unit's baseline score from its follow-up score to get a single column of changes d_i, then divide the average change d̄ by its standard error s_d / √n.

y_Before,i, y_After,i

Matched Pair MeasurementsYOUR DATA ⓘ

Observed metric values for unit i under the two conditions or time points.

d_i

Unit-Level DifferenceOUTPUT ⓘ

Within-pair change for unit i: y_{After,i} − y_{Before,i}.

Number of Matched PairsYOUR DATA ⓘ

Total count of complete paired observations (degrees of freedom df = n − 1).

d̄, s_d

Mean & SD of DifferencesESTIMATED ⓘ

Sample mean and sample standard deviation of the n within-pair differences.

Δ₀

Hypothesized Mean ShiftYOUR DATA ⓘ

Null benchmark difference under H₀ (defaults to 0).

Paired t-StatisticINFERENCE ⓘ

Test statistic compared against Student’s t distribution with df = n − 1.

2. Confidence Interval & Standardized Effect Sizes (Cohen's d_z, Hedges' g_z)

\[ \text{CI}_{1-\alpha} = \bar{d} \pm t_{\alpha/2,\,n-1}\,\frac{s_d}{\sqrt{n}}, \qquad d_z = \frac{\bar{d}}{s_d}, \qquad g_z = d_z \cdot \left(1 - \frac{3}{4(n - 1) - 1}\right) \]

💬 Read it in plain English: The confidence interval gives the plausible range for the true average change, while Cohen's d_z expresses the average shift as a multiple of the standard deviation of changes.

t_{α/2, n−1}

Critical t-ValueINFERENCE ⓘ

Two-tailed critical value from Student’s t distribution at significance level α.

CI_1−α

Confidence Interval for d̄INFERENCE ⓘ

Plausible range for the population mean within-pair difference.

d_z

Cohen’s d_z (Within-Subject)OUTPUT ⓘ

Standardized mean change relative to the variability of pair differences.

g_z

Hedges’ g_z (Bias-Corrected)OUTPUT ⓘ

Small-sample corrected standardized effect size for paired designs.

DIAGNOSTICS & ASSUMPTIONS

Assumption checks

The paired t-test assumes the differences follow an approximately normal distribution, pairs are matched correctly, and each pair is independent of the others. We will summarize diagnostics here after you provide data.