本文針對交易價格為非線性函數之農產品破壞性檢驗進行研究，在考量檢驗成本及因抽樣誤差所造成之誤判損失成本下，以貝氏估計法對母體不良率進行估計，並建構一總損失成本函數，且經電腦數值模擬分析，找出使總損失成本最小化之最適抽樣個數。文中並針對單位檢驗成本、單位價格上限、可容許不良率上限、母體不良率估計值、交易貨品總重量及平均單位重量等進行敏感度分析，並獲得七項具體結論：1.在非線性價格函數下，總損失成本函數 為n之凸函數(convex function)，因此可經由數值分析方法，找出最適抽樣個數n*，使總損失成本最小化。2.在其他條件不變下，單位檢驗成本c增加，最適檢驗個數n*將減少，而總損失成本 將增加。3. 在其他條件不變下，當產品單位價格上限U增加時，最適檢驗個數n*將增加，且總損失成本 亦隨之增加。4. 在其他條件不變下，當可容忍之不良率上限 上升時，最適檢驗個數n*與總損失成本 皆呈先升後降之走勢。5. 在其他條件不變下，當母體不良率估計值 愈接近可容許不良率上限 時，最適破壞性檢驗個數n*及總損失成本 皆趨於最小化。6.在其他條件不變下，當母體總重量 愈重，最適破壞性檢驗個數n*及總損失成本 皆隨之增加。7.在其他條件不變下，每件產品之平均重量w愈重，最適破壞性檢驗個數n*緩慢增加，總損失成本 則緩慢減少。

In this paper, we focus our attention on sample size for destructive inspection of agricultural products under nonlinear price function. Because of the consideration for inspection cost and costs of sampling error(including cost of producer risk and cost of consumer risk), we formulated a mathematical model for total losses. Applying the computerized numerical analysis method, we can find out the optimal sample size that minimizes the total losses. Owing to the fraction rejected P is unknown, we applied Bayesian estimation method to estimate the value of P. Furthermore, sensitivity analysis is taken for unit cost of inspection, upper price per unit, the tolerance limit of fraction rejected, the estimate of population fraction rejected, the total weight of lot and the average weight per unit, respectively. Finally, seven conclusions are drawn for future studies and applications: 1. Total loss function is a convex function on n, so we can apply numerical analysis method to find out the optimal sampling size n* to minimizing total loss cost, 2. Under other’s factors unchanged, when inspection cost increasing, the optimal sampling size n* is decreased and the total loss cost is increased, 3. Under other’s factors unchanged, when the ceiling price increasing, the optimal sampling size n* and the total loss cost are increased, 4. Under other’s factors unchanged, when the upper bound of non-conforming ratio increasing, the optimal sampling size n* and the total loss cost are increased first and decreased after, 5. Under other’s factors unchanged, when the estimate of non-conforming ratio nearing to , the optimal sampling size n* and the total loss cost are decreased, 6. Under other’s factors unchanged, when the total weight increasing, the optimal sampling size n* and the total loss cost are increased, 7. Under other’s factors unchanged, when the average weight per unit w increasing, the optimal sampling size n* is increased slowly and the total loss cost are decreased slightly.