How do I check normality in SPSS?

Checking Normality in SPSS

Many common statistical tests assume that your data, or the residuals of your model, are approximately normally distributed. Before you run those tests, it is good practice to check. In this blog, we'll show how to check normality in SPSS using both plots and statistics, and how to interpret what you find.

Why Normality Matters:

  • Parametric tests such as t-tests, ANOVA and Pearson correlation are based on assumptions about normality.
  • If data are strongly non-normal, especially in small samples, results from these tests may be misleading.
  • Checking normality helps you choose between parametric and non-parametric tests.

Running the Checks in SPSS:

  • Go to Analyze > Descriptive Statistics > Explore.
  • Move your continuous variable into the Dependent List. If you are comparing groups, move the grouping variable into the Factor List so each group is checked separately.
  • Click Plots, tick Histogram and Normality plots with tests, then click Continue and OK.

Reading the Plots:

  • Histogram: look for a roughly bell-shaped, symmetrical distribution.
  • Normal Q-Q plot: if the data are normal, the points will lie close to the diagonal line. Systematic curves away from the line suggest skewness.
  • Boxplot: shows the spread and highlights potential outliers.

Reading the Statistics:

  • The Tests of Normality table shows the Kolmogorov-Smirnov and Shapiro-Wilk tests. Shapiro-Wilk is often preferred for smaller samples.
  • A significant result (commonly p < .05) suggests the data differ from a normal distribution.
  • The Descriptives table shows skewness and kurtosis. Values close to zero suggest a more normal shape. Different sources suggest different cut-offs, so cite the guideline you use.

Interpreting Sensibly:

  • With large samples, formal tests can flag very small, unimportant departures from normality, so always look at the plots as well.
  • With small samples, the tests may lack power to detect real problems.
  • If data are clearly non-normal, consider a non-parametric test, a transformation, or methods that are robust to non-normality.
  • For regression, check the normality of the residuals rather than the raw variables.

Checking normality in SPSS takes only a few clicks. Use plots and statistics together, interpret them in light of your sample size, and report what you checked and what you decided. This will make your choice of statistical test clear and defensible.