Table 2.

Quality assurance data.

Measure	SCZ (n=47)	CON (n=56)	Group Effect	Site Effect	Group × Site Interaction
Absolute Motion	.419 (.0399)	.389 (.0354)	F(1,93102) = 0.32, p=.57	F(4,93102) = 5.81 p<.001	F(4,93102) = 0.42, p=.80
Relative Motion	.127 (.0089)	.102 (.0079)	F(1,93102) = 4.32, p=.041	F(4,93102) = 3.27, p=.015	F(4,93102) = 0.67, p=.62
tSNR	207.6 (6.96)	226.4 (6.18)	F(1,93102) = 4.08, p=.046	F(4,93102) = 6.92, p<.001	F(4,93102) = 2.03, p=.096
Smoothness	10.25 (.120)	10.30 (.106)	F(1,93102) = 0.11, p=.75	F(4,93102) = 6.46, p<.001	F(4,93102) = 0.80, p=.53

Values for SCZ and CON are the estimated marginal means (and SE), controlling for site. Motion and smoothness measures are in mm. Absolute motion is computed relative to a fixed reference frame (middle time point of the run) and relative motion is computed from one frame to the next. Both are outputs of FSL's MCFLIRT tool for motion correction. Smoothness is the full-width half-maximum estimate of the spatial smoothness of the residuals from the first (subject) level GLM (obtained by converting the “resels” estimate from FSL's ‘smoothest’ function into units of mm). Temporal signal-to-noise (tSNR) for each run was obtained by first computing a spatial (voxel-wise) map of mean signal over time divided by the standard deviation of the residuals from the first level GLM. This tSNR map was then averaged over space using a weighted average according to the probability of gray matter at that voxel in MNI152 space. For all measures, values were averaged across the 3 runs for each subject and those subject-specific averages formed the basis for analysis of site and group effects.