So in this method, the regression coefficient is estimated as distribution. This approach considers distribution of the damage parameter. Thus probability of failure evaluation is possible. In addition, the Bayesian estimation is used. This is determined the distribution of the estimated value by against damage true. Further, to adapt to the individual is small sample of, by using the Bayes' theorem, the prior distribution and other individuals. Posterior distribution can be drawn by updating prior information using measurement data. This system is possible to evaluation at small sample, Low cost. When the maintenance is considered from the result of damage evaluation, damage evaluation is not the whole, to estimate a part carefully. So, we propose the "weighting". In this paper, one cases as the delamination identification problem of CFRP laminated plate. Object is the identification of damage in a small sample of large estimation error. Evaluation of the effect of the change of the regression coefficient distribution has on the output result. Influence of an estimated error is different to overestimation or underestimation of probability of failure. The weight function performed influence examination given to a risk value. We investigate relates effective weighting to reduce the risk.