Which statistic is recommended for measuring linear association between two tests and why?

• A. Pearson correlation coefficient, because it accounts for scale differences.
• B. Concordance correlation coefficient, because Pearson ignores scales of the two sets.
• C. Intraclass correlation coefficient, because it is unaffected by systematic bias.
• D. Limits of agreement plots, because they show agreement.

Answer: (B)

When comparing two tests that aim to measure the same thing, you want to know not just that they move together, but that they produce similar values across the range. Pearson correlation captures how tightly the scores co-vary along a straight line, but it can be high even if one test consistently gives higher values than the other or uses a different scale. The concordance correlation coefficient blends the strength of the linear relationship with a measure of how close the scores are to the line of equality (y = x). It penalizes both lack of precision (scatter around the best-fit line) and systematic bias (differences in scale or mean between the tests). This makes it a single statistic that reflects both association and agreement across the measurement range. So, it’s the best choice when you want a summary of how well two tests agree in their numeric values, not just how well they correlate.