Recent studies have shown a high interest in statistical methods dedicated to the prediction of worst case scenarios in Electromagnetic Compatibility. In particular, it has been shown that the extremal types theorem allows for avoiding the use of the statistical terminal points and safe margins in the design of protective devices. Nevertheless, its application requires a large set of measured samples, which is known to be time-consuming in practical experiments. This paper deals with this issue, thanks to a bootstrapping procedure. It is shown that the latter procedure enables to significantly reduce the number of experiments.