This guide explains how to read the p-value from a normality test (like Shapiro-Wilk, Anderson-Darling, or Kolmogorov-Smirnov) to decide if your data follows a normal distribution. You input your test statistic and degrees of freedom or sample size, and the calculator gives you the p-value. Then you compare it to your significance level (usually 0.05) to make a decision.
Formula
For a normality test, the p-value is the probability of observing a test statistic as extreme as, or more extreme than, the one computed from your sample, assuming the data truly come from a normal distribution. The exact formula depends on the test. For the Shapiro-Wilk test, the p-value is often approximated using the transformation:
p ≈ exp( -0.7 × (1 + 0.4 × (n − 3)) × (1 − W)² )
Where:
- W = Shapiro-Wilk test statistic (unitless, between 0 and 1).
- n = sample size (number of observations).
- p = p-value (probability, between 0 and 1).
This is an approximation; statistical software uses more precise algorithms. The calculator above uses this approximation to give you a quick estimate.
Worked Example
Suppose you run a Shapiro-Wilk test on a sample of 30 observations and get W = 0.95. Using the default values from the calculator:
p = exp( -0.7 × (1 + 0.4 × (30 − 3)) × (1 − 0.95)² )
First, compute (30 − 3) = 27. Then 0.4 × 27 = 10.8. Add 1: 11.8. Now (1 − 0.95)² = 0.05² = 0.0025. Multiply: 0.7 × 11.8 × 0.0025 = 0.02065. Finally, exp(−0.02065) ≈ 0.9796. So the p-value is approximately 0.98.
Since 0.98 > 0.05, you fail to reject the null hypothesis of normality. This means your data do not show significant evidence of non-normality.
Assumptions
- The approximation formula is valid for the Shapiro-Wilk test; other tests have different formulas.
- The sample is a random sample from the population.
- The test statistic W is correctly computed from the data.
- The p-value is two-sided (most normality tests report two-sided p-values).
How to Interpret the P-Value
In a normality test, the null hypothesis (H₀) states that the data come from a normal distribution. The alternative hypothesis (H₁) states that the data do not come from a normal distribution. The p-value tells you the probability of getting a test statistic as extreme as yours if H₀ is true.
- If p-value ≥ α (e.g., 0.05): Fail to reject H₀. You do not have sufficient evidence to claim the data are non-normal. In practice, you can proceed with methods that assume normality.
- If p-value < α: Reject H₀. You have evidence that the data are not normally distributed. Consider transformations (e.g., log) or non-parametric tests.
Common significance levels are 0.05, 0.01, and 0.10. Choose α before running the test to avoid bias.
Common Mistakes
- Confusing p-value with the probability that the data are normal. The p-value is not the probability that H₀ is true; it’s the probability of the data given H₀.
- Using the wrong significance level. Always decide α beforehand.
- Ignoring sample size. With large samples (n > 1000), even trivial deviations become significant. With small samples (n < 10), the test may lack power.
- Misreading the test statistic. For Shapiro-Wilk, W close to 1 indicates normality; for Anderson-Darling, smaller values are better.
- Using the p-value alone. Always complement with a histogram or Q-Q plot.
Tips
- Use the p-value as a guide, not an absolute truth. Visual inspection is also important.
- For small samples (n < 30), normality tests are less reliable; consider using the Shapiro-Wilk test, which is more powerful.
- If your data are clearly skewed (e.g., income, reaction times), expect a low p-value and plan to use non-parametric methods.
- When in doubt, run multiple normality tests (Shapiro-Wilk, Anderson-Darling) and see if they agree.
- Remember: failing to reject normality does not prove normality; it just means the data are consistent with normality.
Troubleshooting
My p-value is 0.0000. This often happens with large sample sizes or extreme non-normality. Report as p < 0.001.
My p-value is exactly 1. This can occur with very small samples or when the test statistic is exactly 1. Check your data for ties or constant values.
My p-value from the calculator differs from software. The calculator uses an approximation; software uses exact methods. The difference is usually small, but for critical decisions, rely on the software output.
My data are clearly normal but p < 0.05. This can happen with large samples; the test detects tiny deviations. Look at the effect size or use a Q-Q plot.
Frequently Asked Questions
Q: What is a good p-value for normality?
A: A p-value above your significance level (commonly 0.05) is considered good, meaning you fail to reject normality.
Q: Can I use parametric tests if p > 0.05?
A: Yes, if the p-value is above 0.05, you do not have evidence against normality, so parametric tests are generally appropriate.
Q: What if my p-value is exactly 0.05?
A: It’s on the boundary. Some researchers treat it as significant, others not. Be consistent and report the exact p-value.
Q: Does sample size affect the p-value?
A: Yes. Larger samples make the test more sensitive to small deviations, so p-values tend to be smaller even for minor non-normality.
Q: Which normality test should I use?
A: Shapiro-Wilk is recommended for most cases, especially with n < 50. Anderson-Darling is also good. For very large samples, even small deviations matter, so consider visual checks.
Q: Can I still use t-tests if my data are not normal?
A: For large samples (n > 30), the Central Limit Theorem makes t-tests robust to non-normality. For small samples, consider non-parametric alternatives like Mann-Whitney U.
Related
Formula used
in m | Test statistic (W or D) | - | 0.95 | 0 | 1 in n | Sample size | - | 30 | 3 | 10000 out p | P-value (two-sided) | - | exp(-0.7 * (1 + 0.4 * (n - 3)) * (1 - m)^2) | 4
Each line is one input (in, sel, list) or one result (out: label, unit, formula, decimals). How the calculators work
Method: the standard formula for this calculation with the constants and conversion factors written on the page, rounded as each result states. See method and constants, the disclaimer or report an error.
More in this section
- Normality to g/L Converter & Formula29.220 g/L
- Normality to meq/mL Conversion Calculator1.0000 meq/mL
- Normality Test Q-Q Plot: How to Read It
- Normality Test Results: How to Analyze Them
- Normality Slay the Spire?
- Normality Test for Paired t Test: How to Check