Figure 8.

Distribution of power estimates based on the null hypothesis that the scanner type makes no difference. Here we computed minimum sample size estimates based on 10,000 random permutations mixing 1.5 and 3 T scans for TBM analysis (top row) and for SIENA analysis (bottom row). For TBM, the “power estimates” on the x‐axis represent the estimated sample sizes (AD: n80 = 37 for 1.5 T and 49 for 3 T; MCI: n80=107 for 1.5 T and 166 for 3 T) required to detect a 25% reduction in atrophy (with 80% power) for AD (left) and MCI (right) groups. For SIENA, the estimated sample sizes are numerically greater (AD: n80 = 116 for 1.5 T and 92 for 3 T; MCI: n80 = 207 for 1.5 T and 265 for 3 T). The red lines indicate the estimated sample sizes for designs that exclusively used 1.5 T scans or exclusively used 3 T; dashed lines indicate the boundary that excludes the outer 5% of the null distribution of sample sizes that occur from randomly mixing scanners. None of the red lines falls in the outer 5% of the null distribution, so we cannot reject the null hypothesis that the scanner field strength makes no difference. In other words, there is no detectable difference in power; power was excellent at both field strengths, and, as expected, was somewhat higher for AD than for MCI. For TBM analysis, the 3 T scanner gave a power estimate, for the MCI group, that was within the outer 11% of the power estimates obtained by randomly mixing scanners. The 3 T scanner gave a power estimate, for the AD group, that was within the outer 25% of the power estimates obtained by randomly mixing scanners. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]