In the test evaluation for small sample size, the prior information is generally fused to obtain the evaluation result with weight calculated by the data consistency check for test information and prior information. However, in small sample case, the consistency between two kinds of samples may be unstable. To improve this situation, the physical equivalency credibility is defined by analyzing different test surroundings. Moreover, the composite equivalency weight of prior sample is decided by fusing physical equivalency credibility and data consistency. A weighted method for considering the credibility of the prior information is proposed for Bayesian estimation algorithm and the weighted estimation of normal inverse-Gamma distribution parameters is provided. The efficiency of the prior sample is analyzed by comparing the posterior variances for different cases, and is further applied to determine whether the prior samples should be fused in the conception that only those prior samples which can reduce the posterior variance should be used. Theoretical analysis and simulation demonstrate that this method is credible.