The thesis develops a Bayesian decision model for making an optimum selection among suppliers who provide raw materials of non-repairable products. The model takes into account sampling information, inspection errors, inspection cost, material purchase cost, and product failure cost, with minimizing the total cost as the model objective. Two types of inspection errors are considered in our model. One is the error due to taking a good component for a bad one, and the other is the error due to taking a bad component for a good one. The research first applies the theory of decision trees to construct the total objective functions of the model. Then all formulas relevant to the model are derived and a computer program written by C++ programming language is developed according to the formulas and is included in the appendix of the thesis. The computer program can execute the computation in several minutes on the model with a lot size of 2000 units. The research also performs experimental analysis on the model to show how the component quality, the inspection errors, product failure cost, and purchase cost have impact on the optimum total cost and the optimum sample size.