Abstract: The statistical results of the reliability growth test of multi systems have been investigated. The results show that some useful information would be lost if the synchronizing correction is used in the test, and because a multi sample statistical test will become a single sample test, whose failure probability obeys the combined distribution, the estimation results would not be correct. The synchronizing correction reliability growth test of multi systems is not better than the asynchronous one in providing estimation information, and it will spoil the statistical test. So, it is reasonable to use the asynchronous reliability growth test rather than the synchronous one.

Key words: reliability grow th, mult ip le-system s asynch ronous-co rrect ion model, mult ip le-system s synch roco rrect ion model, stat istacs test