Statistics homework is hard enough. Guessing whether your data is “normal” should not be.
Before you use the calculator: This tool helps you test whether your dataset is consistent with a normal (bell-shaped) distribution. You will get p-values (decision-friendly) plus a visual check (Q-Q plot). Use it to choose the right method (like t-tests/ANOVA vs nonparametric tests) and to write a clean “assumptions” sentence in your paper.
What this calculator does
This normality test calculator takes one list of numbers and returns normality evidence in two ways:
- Formal tests (p-values)
- Shapiro-Wilk (recommended default in many classes)
- Anderson-Darling (tail-sensitive)
- Kolmogorov-Smirnov style checks (see K-S comparison section)
- Jarque-Bera (uses skewness and kurtosis)
- Visual check (Q-Q plot)
- Dots close to a straight line usually mean “closer to normal.”
- Big curves or far-off dots often mean skewness or outliers.
What this means
A normality test checks whether your data are consistent with a normal (bell-shaped) distribution. If p-value < alpha, the test suggests your data are not normal. If p-value ≥ alpha, your data look normal enough for many classroom methods.
In writeups: “Normality was assessed using JB/AD/Lilliefors/KS/K² tests (α = 0.05).”
- Large n can reject normality even for tiny departures.
- Use the Q–Q plot to check shape and outliers.
- Many common tests are robust when data are roughly symmetric.
Show steps / formulas
Step 1: Compute mean and SD, then z-scores: z = (x − mean) / SD.
Jarque–Bera: JB = (n/6)[S² + (K−3)²/4], where S is skewness and K is kurtosis. p-value from χ²(df=2).
Anderson–Darling: compares the sample CDF vs the normal CDF (more weight in the tails). Uses a corrected A²* and an approximation for p-value.
Lilliefors (KS): D is the max gap between empirical CDF and normal CDF when mean/SD are estimated from the sample.
Kolmogorov–Smirnov (KS): Classic KS assumes the normal distribution is fully specified (you know μ and σ). If you estimate μ and σ from the same sample, use Lilliefors instead.
D’Agostino–Pearson: K² combines skewness and kurtosis into one statistic (omnibus). This calculator uses a practical classroom approximation.
Common mistakes
- Pasting non-numeric entries (like “12%”).
- Using too few data points.
- Thinking “p ≥ 0.05 proves normality.”
- Ignoring outliers.
- Using classic KS without specifying μ and σ (use Lilliefors instead).
Need probability or percentiles after you confirm a normal model? Use the Normal Distribution Calculator.
When to use it (and when not to)
Use it when
- Your next step is a method that assumes normality (examples: t-tests, ANOVA, regression residual checks).
- Your teacher asks for “Check normality” or “Test assumption of normality.”
- You have raw data (a list of values), not just summary stats.
Do not overuse it when
- n is huge: tests can reject normality for tiny, unimportant wiggles. Use the Q-Q plot and practical judgment too.
- Your variable is clearly not continuous (for example, Likert totals with very few levels can be tricky; counts often need different models).
- The decision is already clear from context (example: extreme outliers, impossible symmetry).
How it works (simple explanation)
All normality tests start with the same idea:
- Null hypothesis (H₀): the data come from a normal distribution.
- If p-value < alpha (often 0.05): reject normality (data look not normal).
- If p-value ≥ alpha: you do not reject normality (data look normal enough for many classroom uses).
Shapiro-Wilk (default section)
Shapiro-Wilk checks how closely your ordered data match what “normal” data should look like. It is widely recommended because it tends to have good power in many settings.
K-S comparison (section)
Kolmogorov-Smirnov (K-S) compares your sample’s cumulative pattern to a target distribution’s cumulative pattern. The classic K-S test is “distribution free” for some settings, but when you estimate mean and SD from the same sample, the usual K-S critical values do not apply. That is why many classes use a Lilliefors correction (a K-S variant for estimated mean/SD).
Practical takeaway:
- If someone says “K-S for normality,” ask: classic K-S or Lilliefors (K-S with estimated mean/SD)?
- Different software defaults can explain different p-values.
Step-by-step example
Dataset (n = 10):12, 13, 13, 14, 14, 15, 15, 16, 18, 22
Step 1: Choose alpha: Use alpha = 0.05 (typical).
Step 2: Run a normality test (example: Shapiro-Wilk): Suppose the output is p = 0.03.
Step 3: Decide: Since 0.03 < 0.05, you would write: “Reject normality (Shapiro-Wilk, α = 0.05).”
Step 4: Do a quick visual check: A Q-Q plot would often show the last value (22) pulling away (possible outlier/right tail).
Interpretation
Use this simple decision rule:
- p-value < alpha: data are likely not normal (at least one test sees a meaningful departure).
- p-value ≥ alpha: data look normal enough for many classroom methods (you did not detect strong evidence against normality).
A clean writeup line you can copy:
“Normality was assessed using the Shapiro-Wilk test (α = 0.05).”
What to do if not normal
If your result suggests “not normal,” you usually have four student-friendly options:
- Check for outliers and data entry issues: One extreme value can break normality.
- Use a robust or nonparametric method: Many nonparametric tests use ranks and don’t require normality.
Related article: Parametric vs nonparametric tests - Transform the data (only if your class allows it): Common examples are log or square-root transforms for right-skewed data.
- Rely more on the plot + context (especially for big n): With large samples, tiny deviations can trigger rejection even when the method still works fine in practice.
Common mistakes
- Using p ≥ 0.05 to “prove” normality. It only means you did not detect strong evidence against it.
- Pasting too few values (normality tests need enough data to be meaningful).
- Forgetting that software may run different tests by default (Shapiro-Wilk vs K-S vs Lilliefors).
- Ignoring the Q-Q plot (tests and plots should agree “most of the time,” but plots explain why).
- Changing alpha after seeing the result (pick alpha first).
Frequently Asked Questions
What does this result mean?
If p-value < alpha, the test suggests your data are not consistent with a normal distribution. If p-value ≥ alpha, your data look normal enough for many classroom methods.
Which test should I use?
For most school and early-college work, Shapiro-Wilk is a solid default. Use the Q-Q plot alongside it.
Is Kolmogorov-Smirnov popular for normality?
It’s common to hear “K-S,” but for normality with mean and SD estimated from the sample, many tools use a Lilliefors-style correction rather than classic K-S.
Why is my answer different from my teacher’s (or another calculator)?
Most differences come from:
-different test choice (Shapiro-Wilk vs AD vs K-S/Lilliefors),
-different approximations for p-values,
-different rounding.
What sample size is “too big” for normality tests?
There is no single cutoff, but with large n, tests can flag tiny departures. Use the Q-Q plot and practical judgment too.
If it’s not normal, do I automatically use nonparametric tests?
Not automatically. First check outliers and context. Then decide between robust methods, transformations, or nonparametric tests based on your lesson and goal.
Do I need normal data for a t-test?
Many t-based methods are fairly robust when data are roughly symmetric and not dominated by extreme outliers, especially with moderate to large samples. Still, your course may require a formal normality check.
Related BrainMattersLearning statistics tools
Start here (hub):
Helpful next tools (sideways/forward):
Category pages (good for browsing tools):
References
Ghasemi, A., & Zahediasl, S. (2012). Normality tests for statistical analysis: A guide for non-statisticians. International Journal of Endocrinology and Metabolism, 10(2), 486–489. https://pmc.ncbi.nlm.nih.gov/articles/PMC3693611/
NIST/SEMATECH. (2012–2023). e-Handbook of Statistical Methods: Anderson-Darling and Shapiro-Wilk tests. National Institute of Standards and Technology. https://www.itl.nist.gov/div898/handbook/prc/section2/prc213.htm
NIST/SEMATECH. (n.d.). Jarque-Bera test. National Institute of Standards and Technology (Dataplot Reference Manual). https://www.itl.nist.gov/div898/software/dataplot/refman1/auxillar/jarqbera.htm
OpenStax. (2023). Introductory Statistics 2e: The standard normal distribution. Rice University. https://openstax.org/books/introductory-statistics-2e/pages/6-1-the-standard-normal-distribution