Which statistical test should I use?
Choosing the wrong test is one of the most common comments thesis committees make. Answer a few questions about what you want to do, the type of variable, whether you compare different people or the same people, and whether your data are normal. We tell you which test to use, its alternative, which effect size to report, and take you to the calculator.
The calculation runs in your browser. We don’t store what you type.
How it chooses
The goal: compare groups, relate variables, predict, or compare with an expected distribution.
The variable type: numeric (total scores), ordinal (a single Likert item, levels) or categorical (yes/no, major).
The design: groups of different people (independent) or the same people measured several times (related).
Normality: numeric normal data use parametric tests, which compare means; otherwise non-parametric tests, which compare ranks.
Quick table
- Two independent groups: t-test or Mann-Whitney U.
- Before and after: paired t or Wilcoxon; with yes/no answers, McNemar.
- Three or more groups: ANOVA or Kruskal-Wallis. Three or more measurements of the same people: repeated-measures ANOVA or Friedman.
- Relationship between two variables: Pearson or Spearman; two categorical variables: chi-square.
In the glossary: p-value · Effect size · Parametric and non-parametric tests
Next stepNormality test (Shapiro-Wilk and Kolmogorov-Smirnov)Don’t know yet whether your data are normal? Test it with Shapiro-Wilk before choosing.Step-by-step guideWhich statistical test to use in your thesis: a decision-tree guideHow to choose your thesis’s statistical test: decision tree, table of parametric and non-parametric tests and common mistakes. With a free test selector.Frequently asked questions
- Can I use parametric tests with Likert scales?
- With the total score of a multi-item scale, and if it’s normal, yes: that’s the most common case. With a single Likert item, which is ordinal, use non-parametric tests.
- What if one group is normal and another isn’t?
- Use the non-parametric test. The parametric test assumes normality in every group.
- What if my sample is large?
- With more than 30 to 50 people per group, the t-test and ANOVA cope well with non-normality. Still, many committees ask for the non-parametric test when the normality test is significant: check your school’s rules.
Know what you want to ask?
Build the survey in SurvIQ, share it by link, WhatsApp or QR code, and watch results come in live. The Free plan is free forever.
Create a free survey