Figure 5.

(a) The power of the t-test for standardised means for the case where correlations ${\rho }_x\neg ={\rho }_0$ are chosen in equation (2) to compute the standardised means ${\overline{Z}}_i$ for the reference scenario. Figure (b) shows the power under the assumption that the effect size follows a mixture distribution with weight $1-p$ on 0 and weight p on a uniform distribution on $[0,2\delta ]$, using the t-test applied to the per-subject means (mean) and standardised per-subject means (Kiresuk), and the test based on the GEE model (GEE) for the reference scenario.