Figure 3.

Consistency as the pilot size is increased. To evaluate whether power estimates obtained from smaller pilots are in fact representative of larger data sets, the calculations were performed with subsets of the Chin et al. data. Resampling without replacement was used to obtain subsets from 10% to 90% of the original data set. Each resampling was repeated ten times and results averaged. The horizontal position of the small symbols mark the size of the subset used to obtain each power estimate.