Research Measurements
I²
I-Squared Statistic
A measure of the percentage of variability across studies due to true differences.
The I-squared (I²) statistic quantifies the proportion of total variability in a meta-analysis that is attributable to heterogeneity (true between-study differences) rather than sampling error. Values range from 0% to 100%: I² < 25% indicates low heterogeneity (studies are consistent), 25-75% moderate, and > 75% high heterogeneity. In supplement research, high I² often reflects differences in dosing (e.g., 1,000 vs 10,000 IU vitamin D), duration (4 weeks vs 12 months), populations (deficient vs sufficient), or outcome measures. Random-effects models are preferred when I² is substantial.