Which Statistical Test Should I Use?

How to choose a statistical test

Picking the right test depends on three things: what you want to know, how many groups or variables you have, and what type of data you collected. Choosing the wrong test can lead to misleading p-values, so it is worth getting this step right before running any analysis. Once you have chosen a test, our hypothesis testing service can run it for you.

Step 1: define the question

• Compare groups: is the average, median or proportion different between groups or conditions?
• Relationship: do two variables move together?
• Compare with a value: does one group differ from a benchmark?
• Prediction: which variables explain an outcome?

Step 2: identify the outcome type

Numeric data that are roughly bell-shaped usually call for parametric tests such as the t-test, ANOVA and Pearson correlation. Skewed numeric data and ordinal scales, such as a five-point Likert item, usually call for non-parametric tests such as Mann-Whitney U, Kruskal-Wallis and Spearman correlation. Categorical outcomes call for chi-square tests or logistic regression.

Step 3: check independence

Ask whether each observation comes from a different person. If the same people are measured twice, or participants are matched, you need paired or repeated-measures versions of the tests.

Common examples

• Two independent groups, numeric outcome: independent t-test.
• Same people before and after: paired t-test.
• Three groups, numeric outcome: one-way ANOVA.
• Two categorical variables: chi-square test of independence.
• Yes/no outcome with several predictors: logistic regression.

Related analysis services

We can choose and run the right test for your study through hypothesis testing, correlation and regression analysis and analysis in R, Python and Excel. See all services or tell us about your project.