A Monte-Carlo Simulation Package, Multiple Comparison Corrections and Power Estimation Incorporating Secondary Supportive Evidence

Various approaches have been proposed to account for the family-wise type-I errors in neuroimaging studies. This study introduces new global features as alternatives to address the multiple-comparison issue. These global features can serve as alternative brain indices whose type-I error theoretical calculations are unknown. A Monte-Carlo simulation package was used to calculate the family-wise type-I error of the newly introduced global features, as well as the conventional multiple comparison corrected p-values related to the height of the statistic (and cluster size) of interest in situations where random field theorem based p-values might be validated. In addition, this package was designed to perform statistical power analyses, taking multiple comparisons into consideration for the conventional statistics and the new global features. The behaviors of the global index type-I error thresholds as a function of the degrees of freedom (D) of t-distribution were investigated. Data from an oxygen-15 water PET study of right hand movement was used to illustrate the use of the global features and their type-I error and statistical power. With this PET example, we showed the superior statistical power of some global indices in cases where there were moderate changes over a relatively large brain volume. We believe that the global features and the calculation of type-I errors/statistical powers by the computer simulation package provide researchers alternative ways to account for multiple comparisons in neuroimaging studies.