Best statistic for reliability for continuous data (ratio or interval) is which term?

Best statistic for reliability for continuous data (ratio or interval) is which term?

• A. Pearson correlation
• B. Intraclass correlation coefficient
• C. Kappa statistic
• D. Spearman correlation

Answer: (B)

Reliability for continuous data is best assessed with the intraclass correlation coefficient because it directly measures how close repeated measurements or ratings are in actual units for the same subjects. It captures agreement in the observed values, not just whether the measurements move together. Pearson correlation looks at linear association between two sets of measurements, but high association doesn’t guarantee good agreement. You can have measurements that track together but differ by a consistent bias or scale, so correlation alone can overstate reliability. Spearman correlation does the same idea in ranks rather than actual values, so it also misses true agreement. Kappa is designed for categorical data, not continuous scales, so it’s not appropriate for reliability of interval or ratio measurements. ICC can be used with ratio or interval data and can be tailored to different study designs (single vs. average measurements, and whether you’re focusing on consistency or absolute agreement). This makes it the most informative statistic for assessing reliability of continuous measurements across repeated tests or different raters.

Reliability for continuous data is best assessed with the intraclass correlation coefficient because it directly measures how close repeated measurements or ratings are in actual units for the same subjects. It captures agreement in the observed values, not just whether the measurements move together.

Pearson correlation looks at linear association between two sets of measurements, but high association doesn’t guarantee good agreement. You can have measurements that track together but differ by a consistent bias or scale, so correlation alone can overstate reliability. Spearman correlation does the same idea in ranks rather than actual values, so it also misses true agreement.

Kappa is designed for categorical data, not continuous scales, so it’s not appropriate for reliability of interval or ratio measurements.

ICC can be used with ratio or interval data and can be tailored to different study designs (single vs. average measurements, and whether you’re focusing on consistency or absolute agreement). This makes it the most informative statistic for assessing reliability of continuous measurements across repeated tests or different raters.