Paired t-test

Dependent (outcome) variable: Continuous/numerical/scale (some disciplines also test ordinal data)
Independent (predictor/explanatory) variable: Time or condition (two levels only)

Use: A paired t-test requires two measurements of the same variable for each subject. This could be a comparison of measurements taken before and after an intervention or compare how a group of subjects perform under two different test conditions. The test calculates the paired difference for each subject and assesses whether the mean of these paired differences is significantly different to zero.
Example: Test the effectiveness of a margarine for reducing cholesterol by comparing participants cholesterol before a trial to their cholesterol after 4 weeks of using the margarine.

Assumptions: The paired differences should be approximately normally distributed. If a histogram or QQplot suggests the differences are vey skewed or the dependent variable is ordinal, a Wilcoxon signed rank test (which is based on ranks rather than the raw data) should be used. Some people use parametric tests for ordinal data if there are quite a few categories but be careful when interpreting differences

Data notes Ensure there is one row per participant with a different column for the dependent variable for each condition/time point.

Click on the download button to get the appropriate resource for the test and software package you want help with.

	SPSS	R Sheet	R Script	Jamovi	SAS	Mathematical Understanding
Paired t-test
Wiloxon signed rank (non-parametric paired t-test equivalent)

Need help choosing the right statistical test? Use our online test chooser