Resampling (bootstrapping) the mean: A definite do

Abstract

There has been discussion lately about whether bootstrap versus permutations methods are the more appropriate in diagnostic psychophysiology, especially in the context of concealed information tests (CITs), also called guilty knowledge tests (GKTs). We review the use of bootstrapping in most published CIT/GKT studies (where means, or correlations, as opposed to peak amplitudes are estimated) and find them valid and reliable. When bootstrapping and permutations were applied to the same synthesized ERP datasets, they agreed 98.1% of the time, averaged over 24,000 iterations.