Explore (normality, outliers)
Explore describes a scale variable in depth — confidence interval for the mean, trimmed mean, median, IQR, percentiles, the Shapiro-Wilk normality test and boxplots — overall or separately for each group of a factor.
Menu: Analyze → Descriptive Statistics → Explore…
When should I use Explore?
- Checking the normality assumption before a t test, ANOVA or regression.
- Finding outliers (boxplot points are labelled with case numbers).
- Comparing distributions across groups (add a factor).
What are the assumptions?
- Shapiro-Wilk needs 3 to 5,000 valid cases.
How do I run it in StatLab?
- Choose dependent variable(s); optionally a factor such as group.
- Keep Boxplots ticked; add Histograms if wanted.
How do I read the output?
| Output | Meaning |
|---|---|
| 95% CI for Mean | Range that would contain the population mean in 95% of samples. |
| 5% Trimmed Mean | Mean after removing the top and bottom 5% (fractional weighting). Close to the mean → outliers have little influence. |
| Interquartile Range | P75 − P25 (weighted-average percentiles). |
| Tukey's Hinges | Quartiles used to draw the boxplot. |
| Shapiro-Wilk W, df, Sig. | W near 1 = close to normal. Sig. < .05 means the data depart significantly from a normal distribution (Royston 1995 algorithm). |
| Boxplot ○ / * | ○ = outlier (1.5–3 box lengths beyond the box), * = extreme (more than 3). |
Worked example
Explore score in the bundled sample dataset (File → Sample dataset: 150 survey respondents with gender, programme group, age, pre/post test scores, five satisfaction items, income, smoking and education): Shapiro-Wilk W = .990, df = 150, Sig. = .341 — no significant departure from normality.
How would I report it in APA style?
A Shapiro-Wilk test indicated that wellbeing scores did not deviate significantly from being normally distributed, W(150) = .990, p = .341.
Numbers above were computed by StatLab’s validated engine from the sample dataset when this page was built.
What are common mistakes?
- With large samples (n > 300) Shapiro-Wilk flags trivial deviations — look at the histogram and skewness too.
- Testing normality of the whole sample when the test assumes normality within each group: add the factor.