Anderson-Darling Test

Is your data really normal? Find out in one tap.

Anderson-Darling Test is a fast, fully offline tool that checks whether your numbers follow a normal (bell-shaped) distribution. It runs the Anderson-Darling test of normality - the test that weights the TAILS most, so outliers, skew and fat tails get caught where other tests miss them - and gives you a clear, plain-language verdict: keep normality, or reject it.

No accounts. No internet needed. No data leaves your phone.

WHAT YOU GET
- A clear PASS / FAIL verdict with the adjusted A* statistic judged against the 5% critical value (0.787).
- The exact numbers: A-squared (raw), A* (adjusted), the p-value, the widest gap, the mean, the standard deviation and the sample size.
- A p-value from the D'Agostino & Stephens approximation, computed entirely on-device.

VISUALS THAT EXPLAIN THE TEST
- Empirical CDF vs the fitted normal curve: your data's green staircase laid over the grape normal S-curve, with the widest gap marked - the very thing Anderson-Darling measures.
- A* gauge: the adjusted statistic on a track with the 5% critical mark; green when normal, rose when not.
- Q-Q plot and a histogram with the fitted normal curve, for a second look at skew and tails.
All charts are drawn natively - light, smooth and offline.

EASY DATA ENTRY
- Type or paste your values (comma or space separated, thousand-separator friendly).
- In the simulator, set the sample size three ways: type the number, tap the - / + steppers, or drag the slider.

LEARN, NOT JUST CALCULATE
- Idea: what "normally distributed" means and why Anderson-Darling weights the tails, with the formulas rendered offline.
- Sim: pick a shape - bell, uniform, right-skew, heavy tails or bimodal - and watch A* react in real time.
- Cases: worked real datasets (symmetric heights that pass; right-skewed income and an easy exam that fail) so you can see the verdict change.

WHO IT'S FOR
- Students and teachers in statistics, data science and the sciences.
- Researchers, analysts and engineers who need a quick normality check before a t-test, ANOVA or regression.
- Anyone curious about whether the bell-curve assumption actually holds for their data.

THE MATH (TRANSPARENT)
A-squared = -n - (1/n) * sum (2i-1) [ ln Fi + ln(1 - F(n+1-i)) ], with Fi the fitted normal CDF.
A* = A-squared * (1 + 0.75/n + 2.25/n^2), judged against the published critical scale
(0.576 / 0.656 / 0.787 / 0.918 / 1.092). Normal samples give A* near 0.15 (p near 1); a strong
right-skew gives A* near 1.57 (p near 0).

PRIVACY
The app works completely offline. Your data stays on your device. The only network use is to load the ad banner.

Light, accurate and built to teach - download Anderson-Darling Test and see whether your data fits the bell.