Abstract: In the probabilistic safety assessment (PSA) of nuclear power plants, there are few historical records on some initiating event frequencies or component failures in industry. In order to determine the noninformative priors of such reliability parameters and initiating event frequencies, the Jeffreys method in Bayesian statistics was employed. The mathematical mechanism of the Jeffreys prior and the simplified constrained noninformative distribution (SCNID) were elaborated in this paper. The Jeffreys noninformative formulas and the credible intervals of the Gamma-Poisson and Beta-Binomial models were introduced. As an example, the small break loss-of-coolant accident (SLOCA) was employed to show the application of the Jeffreys prior in determining an initiating event frequency. The result shows that the Jeffreys method is an effective method for noninformative prior calculation.