Normality Tests Explained: Shapiro-Wilk and What to Do When Your Data Fails

Your supervisor expects normality testing before parametric tests. Here is what normality means, how to run Shapiro-Wilk, how to interpret the result, and what to do when it is violated.

Before running most parametric tests (t-tests, ANOVA, Pearson correlation, regression), you need to check whether your data is approximately normally distributed. If it is not, you need to use a non-parametric alternative. This is called assumption testing, and your supervisor will expect to see it in your Chapter 4.

What is normality?

A normal distribution is the classic bell curve — symmetrical, with most values clustering around the mean and fewer values at the extremes. Many parametric tests assume your data follows this pattern.

In practice, data never follows a perfect normal distribution. The question is whether the deviation is large enough to matter. That is what normality tests assess.

The Shapiro-Wilk test

The Shapiro-Wilk test is the standard normality test for small to medium samples (n up to 50). It is more powerful than the older Kolmogorov-Smirnov test for detecting non-normality.

Null hypothesis (H0): The data is normally distributed
Alternative hypothesis (H1): The data is NOT normally distributed

If p > .05: fail to reject H0 — normality is not significantly violated, proceed with parametric test.
If p < .05: reject H0 — normality is violated, consider non-parametric alternative.

When to use Kolmogorov-Smirnov instead

For larger samples (n > 50), use the Kolmogorov-Smirnov test. ResearchScope automatically selects the appropriate test based on your sample size.

How to report normality in APA format

"Normality was assessed using the Shapiro-Wilk test. Results indicated that the assumption of normality was met for job satisfaction, W = .971, p = .142, and work performance, W = .963, p = .081."

When normality is violated: "The Shapiro-Wilk test revealed a significant departure from normality for the variable anxiety, W = .821, p = .003. Accordingly, the non-parametric Mann-Whitney U test was used in place of the independent samples t-test."

The Q-Q plot

Alongside the numerical test, use a Q-Q (quantile-quantile) plot for visual confirmation. A Q-Q plot shows your sample data quantiles against theoretical normal quantiles. If points fall on or near the diagonal line, your data is approximately normal. Deviations from the line indicate non-normality.

What to do when normality is violated

Parametric test Non-parametric alternative
Independent t-test Mann-Whitney U
Paired t-test Wilcoxon signed-rank
One-way ANOVA Kruskal-Wallis
Pearson r Spearman rho

ResearchScope applies these alternatives automatically. When normality is violated, it switches to the non-parametric test and notes the reason in plain English — for example: "Normality was violated (W = .821, p = .003). Mann-Whitney U was used instead of the independent samples t-test."

Skewness and kurtosis as supplementary checks

Beyond the formal test, check your skewness and kurtosis values. Values outside ±2 for skewness or ±7 for kurtosis generally indicate meaningful departures from normality. ResearchScope reports both in the descriptive statistics table.

Try it on your own data

Upload your CSV and get a full business intelligence report in 30 seconds. Dashboard is completely free, no account needed.

Analyze my data free