Abstract: The following concepts are explained: the appropriate number of test samples of Bootstrap method is n≥5, and Bootstrap method is not suitable for sample size n=1 or n=2. Due to this reason, two kinds of methods are presented to solve this problem . The first method is based on the simiempirical estimation theory. The virtualy resempling the sample size from n=1 to n=5 following the empirical distribution function and the Bootstrap method are then combined as the second method. The theoretical basis and derivation process of the second method are presented. Several useful discussions are given. An example is used to illustrate the concept and algorithm of these methods.

Key words: Bootstrap, resempling sample, extreme small sample, test evaluation