Abstract: The distortion of prior information and too many prior samples would distort the evaluation results for small sample, which can be improved by introducing the credibility of the prior information. However, most of the measures for the credibility of the prior information are based on data by judging whether the prior samples and the test samples are subjected to the same distribution, which is unreliable when the samples is very few. A measure of credibility from the physics resource of the data is proposed in this article. The rule for fusion estimation considering the credibility is also concluded, based on that the flaws of the existing methods considering the credibility are analyzed. Moreover, taking the estimation of the parameters of the normal distribution for example, a posteriorweighted Bayesian estimation method considering the credibility of the prior information is given. Both theoretical analysis and simulation results show that the proposed method is credible and it is better than the conventional Bayesian method.

Key words: physics credibility, weighted fusion, test evaluation, small sample, Bayesian estimation