Table 1. Summary of Methods.

Method summary of the bootstrap simulation and imputation of real missing cases. A) Variance (SEM) and t-statistic measures were collected from both the MAR and MCAR bootstrap simulations, and pooled from 200 permutations within each voxel for two sample sizes and three levels of simulated-missing data. Bootstrap simulations only included voxels with no missing data (33,323) and pseudo-randomly removed subjects from each permuted dataset to simulate missingness and allow direct comparisons with simulated-complete datasets. B) Imputation of real missing data began with an examination of the subject removal method, by counting the number of voxels that were gained by removing each subject. Each of the other measures was performed on 11,747 voxels that were missing between 1-18 cases, and pooled results were compared (indirectly) to the 33,323 complete data voxels. SEM and t-statistics were also collected for each voxel with real missing subjects.

A. MAR and MCAR Bootstrap Simulations	N = 25	N = 49
Simulated Missing Data Proportion	10%, 30%, 50%	10%, 30%, 50%
Missing Data Approach	# Missing Data Tests	# Missing Data Tests
Simulated-Complete Data (for comparisons)	3 × 200 × 33,323	3 × 200 × 33,323
Available Case Analysis	3 × 200 × 33,323	3 × 200 × 33,323
Mean Replacement	3 × 200 × 33,323	3 × 200 × 33,323
Multiple Imputation	3 × 200 × 33,323	3 × 200 × 33,323
Neighbor Replacement (18 mm)	3 × 200 × 33,323	3 × 200 × 33,323
Neighbor Replacement (24 mm)	3 × 200 × 33,323	3 × 200 × 33,323
Total MAR Analyses:	119,962,800	119,962,800
Total MCAR Analyses:	119,962,800	119,962,800
Total Bootstrap Analyses:	479,851,200

B. Imputation of Real Missing Cases
Missing Data Approach	# Missing Data Tests
Subject Removal	-
Available Case Analysis	11,747
Mean Replacement	11,747
Multiple Imputation	11,747
Neighbor Replacement (18 mm)	11,747
Neighbor Replacement (24 mm)	11,747
Total Analyses:	58,735