Statistical test calculators
Two ways in. If you are holding raw data, start with a test calculator. If your software already gave you a test statistic, go straight to the converter.
Start from your data
Paste numbers or enter counts. These compute the test statistic for you, then report the p-value with an effect size and a confidence interval.
You have two columns of measurements
T-Test Calculator
Compare the average of two groups. Independent, paired, or one-sample.
You have visitors and conversions
A/B Test Significance
Compare two conversion rates with the lift and its confidence interval.
You have a table of counts
Chi-Square Calculator
Test a 2×2 up to 5×5 contingency table, with the expected counts and Cramér’s V.
You have three or more groups
ANOVA Calculator
Compare three or more group means at once, with η² and the per-group summaries.
You know the population σ
Z-Test Calculator
One- and two-sample Z-tests from raw data or from the mean, SD and sample size.
You have paired X and Y values
Correlation Calculator
Pearson r from paired columns, with r² and an interval on the correlation itself.
Estimate rather than test
A p-value tells you whether an effect is detectable. These answer the questions that usually matter more: how big is it, how precisely have you measured it, and how much data do you need?
You want a range, not a verdict
Confidence Interval
For a mean or a proportion. Proportions use the Wilson score interval, not Wald.
You want to know how big it is
Effect Size Calculator
Cohen’s d and Hedges’ g for two groups, with a confidence interval on the estimate.
You have not run the test yet
Sample Size Calculator
How many observations you need, decided before you collect any of them.
You need the threshold, not the p
Critical Value Calculator
The boundary of the rejection region for z, t, χ², F and r, with the decision rule.
Start from a test statistic
Convert a statistic your software already reported into an exact p-value, with a shaded distribution chart showing what that number represents.
You already have a Z-score
Z-Score to P-Value
Standard normal tail probabilities, exact into the far tail.
You already have a t-statistic
T-Score to P-Value
Student t with any degrees of freedom, including fractional Welch df.
You already have a χ² value
Chi-Square to P-Value
Tests of independence and goodness of fit.
You already have an F-statistic
F-Score to P-Value
ANOVA and regression model tests, with both degrees of freedom.
You already have a correlation
Correlation to P-Value
Test a Pearson r against zero, with the variance explained.
Choosing the right test
Most of the difficulty in statistics is picking the test, not running it. Three questions settle it almost every time.
1. What kind of outcome are you measuring?
If it is a number — revenue, blood pressure, time on page — you are in t-test and ANOVA territory. If it is a category — converted or not, passed or failed — you want a proportion test or chi-square. Getting this wrong is the most common error, and it invalidates everything downstream.
2. How many groups?
One group against a fixed value is a one-sample test. Two groups is a t-test or a two-proportion test. Three or more calls for ANOVA, because running every pairwise t-test instead inflates your false-positive rate badly — with five groups there are ten comparisons and roughly a 40% chance of at least one spurious result.
3. Are the observations paired?
If each value in one group corresponds to a specific value in the other — the same person measured twice, matched pairs — use a paired test. Pairing removes between-subject variation and gives you substantially more power. Applying it to data that are not genuinely paired is a serious error in the opposite direction.
| Outcome | Groups | Test |
|---|---|---|
| Number | 1 vs. a value | One-sample t-test |
| Number | 2 independent | Welch's t-test |
| Number | 2 paired | Paired t-test |
| Number | 3 or more | One-way ANOVA |
| Category | 2 rates | Two-proportion Z-test |
| Category | Contingency table | Chi-square |
| Two numbers | Relationship | Pearson correlation |