Normality Tester

Jarque–Bera and Anderson–Darling normality checks for a numeric series or column.

FreeNo accountOn your device

Runs in your browser. How to verify

Loading tool…

About this tool

Check whether a numeric column is plausibly normal. The tool reports skewness, excess kurtosis, the Jarque-Bera statistic with a p-value, and the Anderson-Darling statistic with a plain-language reading. It needs at least eight values.

How to use

1
Step 1
Paste a CSV (or a plain list of numbers) or drop a file.
2
Step 2
Enter the column name if you pasted a table.
3
Step 3
Press Process.

How it works

Jarque-Bera tests whether skewness and kurtosis match a normal distribution: JB = n/6 · (S² + K²/4), compared with a chi-square distribution with 2 degrees of freedom, so p = exp(−JB/2). Anderson-Darling compares the sorted data with the normal CDF and gives more weight to the tails. Skewness and excess kurtosis are shown as sample-adjusted values (the Excel and pandas convention), while Jarque-Bera uses the population moments from its definition. The AD reading uses Stephens' cut-offs on the small-sample-adjusted statistic A* = A²(1 + 0.75/n + 2.25/n²): below 0.787 not rejected at 5%, up to 0.918 rejected at 5%, up to 1.092 rejected at 2.5%, above that rejected at 1%.

JB = (n/6)(S² + K²/4),  p = e^(−JB/2)

Worked example

A right-skewed sample

Input
y: 2.1, 3.9, 6.2, 7.8, 10.1, 12.2, 30, 8.8, 9.5, 7.1
Output
n 10, mean 9.77, skewness 2.304, excess kurtosis 6.376 Jarque-Bera 10.196, p = 0.0061 Anderson-Darling 1.047: Reject normality at 1%
Walkthrough
The value 30 drives the large skewness and kurtosis, so both tests reject normality. A symmetric sample like 4, 5, 5, 6, 6, 6, 7, 7, 8, 5.5 gives p = 0.92 and AD 0.24, so normality is not rejected.

Limitations

  • Fewer than 8 numeric values gives an error.
  • Jarque-Bera relies on large-sample theory, so its p-value is only approximate for small n.
  • No Anderson-Darling p-value is reported, only a reading against Stephens' critical values.
  • Not rejecting normality is not proof the data is normal.

Questions

How do I test for normality?
Paste the column and press Process. Compare the Jarque-Bera p-value and the Anderson-Darling reading with your threshold, such as 0.05.
How many data points do I need?
At least 8 numeric values.
What does a low Jarque-Bera p-value mean?
The data's skewness and kurtosis differ from a normal curve, so normality is rejected.
Can I paste a plain list of numbers?
Yes. A list of numbers works as well as a CSV column.

Features

  • Jarque-Bera with p-value
  • Anderson-Darling
  • Skewness and kurtosis

Related resources

Guides and videos from around the web—opens on YouTube or the publisher’s site.

  • Finding mean, median, and mode

    Khan Academy

    Descriptive stats for quick data summaries.

    Watch on YouTube
  • What Is A RESTful API?

    Traversy Media

    APIs are how most data tools receive payloads.

    Watch on YouTube