Effect of ICC on sample size requirements. As reproducibility decreases, i.e. ICC is lower, a larger sample size is required to detect all 3 effect sizes with 80% power at an α level of 0.05. For example, when the ICC is 0.6, a sample size of 1.5*25= x, 1.5*50= x, and 1.5*100= x is sufficient to detect strong, medium, and weak effects, respectively.