⚗️  Statistical Toolkit

Statistical Hypothesis Testing

A complete toolkit for rigorous statistical inference — from classical t-tests and ANOVA to non-parametric alternatives. Every test computes exact p-values with detailed interpretations.

7Test types
14+Sub-tests
100%In-browser
ExactP-values

✦ Smart Test Selector — Not sure which test? Let us guide you.

What kind of data are you working with?

This determines the family of tests available to you.

📏
Continuous / Numerical
Measurements, scores, concentrations, times…
🗂️
Categorical / Count
Frequencies, proportions, pass/fail counts…
↔️
Comparing Variability
Are two groups equally variable? (variances)
Continuous data › What are you comparing?

What is the goal of your comparison?

Think about how many groups and whether observations are independent or paired.

🎯
One sample vs. a known value
Is the average equal to a target or standard?
⚖️
Two independent groups
Do two unrelated groups have the same mean?
🔗
Paired / Before-After
Same subjects measured twice — is there a change?
📊
Three or more groups
Comparing means across multiple groups simultaneously.
🔧
Non-normal data / outliers
Data is skewed or small samples — no normality assumption.
Categorical data › What are you testing?

What is the goal of your analysis?

📐
Does distribution fit expected?
Goodness-of-fit: observed vs. expected frequencies.
🔀
Are two variables associated?
Independence test on a contingency table.
📈
Comparing proportions
Is a proportion equal to a target, or are two proportions equal?
Variance comparison ›

How many groups are you comparing?

📉
Two groups
F-test (Variance Ratio Test)
📊
Three or more groups
Use Levene's test (via ANOVA on residuals)
Recommendation

✅ One-Sample t-test

Compare the mean of your sample to a hypothesized value μ₀. The t-test is appropriate when σ is unknown (which is almost always). For known σ and large n, a Z-test is equivalent.

Open t-test Tool →

Also consider: Z-test if population σ is known.

Recommendation

✅ Two-Sample Independent t-test (Welch's)

Compare the means of two independent groups. Welch's t-test is recommended as it does not assume equal variances, making it more robust for most situations.

Open t-test Tool →

If data is clearly non-normal: Mann-Whitney U Test

Recommendation

✅ Paired t-test (Dependent Samples)

When each observation in group 1 is matched to one in group 2 (same subject before/after, matched pairs). The paired t-test analyses the differences directly, reducing within-subject variability.

Open t-test Tool →
Recommendation

✅ One-Way ANOVA (or Two-Way ANOVA)

ANOVA (Analysis of Variance) tests whether at least one group mean differs from the others. Use one-way for a single factor, two-way when you have two categorical factors. Post-hoc tests identify which specific pairs differ.

Open ANOVA Tool →

Non-parametric alternative: Kruskal-Wallis Test

Recommendation

✅ Non-Parametric Tests

When normality cannot be assumed — use rank-based tests that make no distribution assumptions. These are also robust to outliers and suitable for ordinal data.

Mann-Whitney U (2 groups) →   Kruskal-Wallis (3+ groups) →
Recommendation

✅ Chi-Square Goodness-of-Fit Test

Tests whether observed frequencies in categories match expected theoretical frequencies. Suitable for testing if data follows a uniform, Poisson, or any specified distribution.

Open Chi-Square Tool →
Recommendation

✅ Chi-Square Test of Independence

Tests whether two categorical variables are independent in a contingency table. Also quantifies the strength of association with Cramér's V.

Open Chi-Square Tool →
Recommendation

✅ Z-Test for Proportions

Test whether a proportion equals a target value, or compare two proportions from independent samples. Requires n·p ≥ 5 and n·(1-p) ≥ 5 for the normal approximation to hold.

Open Z-Test Tool →
Recommendation

✅ F-Test (Variance Ratio Test)

Tests whether two populations have equal variances. Commonly used as a preliminary test before a pooled two-sample t-test, or to assess process consistency.

Open F-Test Tool →
Recommendation

✅ Homogeneity of Variance

For 3+ groups, the Kruskal-Wallis tool includes a Levene's-style variance comparison. Alternatively, run pairwise F-tests between groups.

Open F-Test Tool →

All Available Tests

Click any card to open the tool directly.

📐
Parametric
Student's t-Test
One-sample · Two-sample · Paired

Compare means when the population standard deviation is unknown. Includes Welch's correction for unequal variances. Computes exact p-values, confidence intervals, and effect sizes.

mean comparison before/after Cohen's d CI
→
📊
Parametric
ANOVA
One-Way · Two-Way

Analysis of Variance for 3+ group comparisons. Full ANOVA table with SS, MS, F-statistic, and p-value. Tukey HSD post-hoc tests identify which group pairs differ significantly.

3+ groups post-hoc interaction effects η²
→
🗂️
Categorical
Chi-Square Test
Goodness of Fit · Independence

Tests for categorical data: fit to a theoretical distribution, or test independence between two categorical variables in a contingency table. Reports Cramér's V effect size.

frequencies contingency table Cramér's V
→
📈
Parametric
Z-Test
Mean · One-proportion · Two-proportions

Z-tests for large samples or known population σ. Covers one-sample mean tests, one-proportion tests against a target, and two-sample proportion comparisons.

proportions large n known σ
→
↔️
Variance
F-Test
Variance Ratio Test

Tests equality of variances between two groups using the F-distribution. Essential for checking the equal-variance assumption before pooled t-tests, and for process capability studies.

variance comparison homoscedasticity process control
→
🔧
Non-Parametric
Mann-Whitney U Test
Wilcoxon Rank-Sum Test

Non-parametric alternative to the two-sample t-test. Works for ordinal data, skewed distributions, and small samples. No normality assumption required.

ordinal data no normality rank-based robust
→
🏆
Non-Parametric
Kruskal-Wallis Test
Non-Parametric One-Way ANOVA

Non-parametric alternative to one-way ANOVA for comparing 3+ independent groups. Based on ranks; no normality required. Includes Dunn's post-hoc test for pairwise comparisons.

3+ groups no normality Dunn's post-hoc
→

Quick Reference: Which test do I need?

A summary with real-world examples to help you choose the right test for your situation.

Situation Example Assumptions Test Link
📏 Comparing Means (Continuous Data)
1 sample vs. a known/target value, σ unknown A factory claims its bags weigh 500g on average. You weigh 25 bags — is the true mean actually 500g? Population ≈ normal, or n ≥ 30 (CLT) One-Sample t-test → Open
1 sample vs. a known/target value, σ known IQ scores are known to have population σ = 15. Is your sample's mean IQ different from 100? Population σ known; normal or large n One-Sample Z-test → Open
2 independent groups, means Do students in Class A score differently on average than students in Class B on the same exam? Approx. normal or n ≥ 30 per group; variances may differ Welch's t-test → Open
Paired / before-after (same subjects) Measure each patient's blood pressure before and after a new medication — did it change? Differences ≈ normally distributed Paired t-test → Open
3+ independent groups, means, 1 factor Compare average crop yield across 4 different fertilizer brands. Normal residuals, roughly equal variances across groups One-Way ANOVA → Open
2 factors at once (+ their interaction) Does yield depend on both fertilizer type AND watering frequency, and do they interact? Normal residuals, balanced design (equal n per cell) Two-Way ANOVA → Open
🗂️ Categorical / Count Data
Observed frequencies vs. an expected distribution Roll a die 120 times — do the 6 outcomes appear equally often, as a fair die would predict? Expected count ≥ 5 in each category Chi-Square Goodness-of-Fit → Open
Two categorical variables — are they related? Is a customer's preferred product (A/B/C) associated with their gender? Expected count ≥ 5 in each cell of the table Chi-Square Independence → Open
Proportion vs. a target value A process should have a ≤5% defect rate. Out of 200 units, 14 were defective — is that significantly higher? n·p₀ ≥ 5 and n·(1−p₀) ≥ 5 Z-Test (one proportion) → Open
Two independent proportions compared Landing page A converts 12% of visitors, page B converts 15% — is that difference real or just noise? n·p̂ ≥ 5 and n·(1−p̂) ≥ 5 in both groups Two-Proportion Z-Test → Open
↔️ Comparing Variability (Spread)
Equality of two variances Machine A and Machine B both fill bottles to 500ml — is one machine more inconsistent (higher variance) than the other? Both samples ≈ normally distributed F-Test → Open
🔧 Non-Parametric (No Normality Required)
2 groups, skewed data / outliers / ordinal scale Compare customer satisfaction ratings (1–5 scale) between two stores — ratings are ordinal, not normal. Independent samples; no distribution shape assumed Mann-Whitney U → Open
3+ groups, skewed data / outliers / ordinal scale Compare repair times (heavily right-skewed) across 3 different technicians. Independent samples; no distribution shape assumed Kruskal-Wallis → Open