For comparing two related means with continuous data (pre/post), which test is appropriate?

For comparing two related means with continuous data (pre/post), which test is appropriate?

• A. Independent t-test
• B. Paired t-test
• C. Chi-square
• D. Mann-Whitney U

Answer: (B)

When the data come from the same subjects measured before and after an intervention, the observations are paired. You want to know if there is a meaningful average change across those pairs. The paired t-test does exactly that by focusing on the differences within each subject and testing whether the mean difference is different from zero. This within-subject comparison reduces variability from individual differences and generally provides more statistical power than treating the two time points as independent. Key assumptions: the differences between paired observations should be approximately normally distributed, and the measurements should be on a continuous (interval or ratio) scale. If the paired differences aren’t normally distributed and the sample is small, a nonparametric alternative is the Wilcoxon signed-rank test, which doesn’t assume normality. Why the other options aren’t appropriate here: the independent t-test compares means between two separate groups with no pairing, which ignores the pre/post relationship. Chi-square tests are for categorical data, not continuous measurements of a mean. Mann-Whitney U compares medians between two independent samples, again not accounting for pairing.

When the data come from the same subjects measured before and after an intervention, the observations are paired. You want to know if there is a meaningful average change across those pairs. The paired t-test does exactly that by focusing on the differences within each subject and testing whether the mean difference is different from zero. This within-subject comparison reduces variability from individual differences and generally provides more statistical power than treating the two time points as independent.

Key assumptions: the differences between paired observations should be approximately normally distributed, and the measurements should be on a continuous (interval or ratio) scale.

If the paired differences aren’t normally distributed and the sample is small, a nonparametric alternative is the Wilcoxon signed-rank test, which doesn’t assume normality.

Why the other options aren’t appropriate here: the independent t-test compares means between two separate groups with no pairing, which ignores the pre/post relationship. Chi-square tests are for categorical data, not continuous measurements of a mean. Mann-Whitney U compares medians between two independent samples, again not accounting for pairing.