在預測犯罪時，經常會使用危險評估量表，而危險評估量表所使用的預測效度（又稱預測準確度）一般多是使用皮爾遜氏積差相關係數(Pearson's r)與接收操作特徵曲線（Receiver Operative Characteristic Curve，簡稱ROC）。然而，何謂ROC？ROC值所代表的意義爲何？其在預測2×2表上與預測之敏感度與特異度有何關聯？犯罪防治人員應如何決定使用較好的量表？此均爲本研究之目的。從發展本土化的婚姻暴力危險評估量表及性侵害加害人靜態量表中發現ROC曲線值在預測2×2表（即是猜測會否再犯及實際是否再犯之表）上似乎是敏感度及特異度之平均值（即二者相加除以二），經由本研究以公式及圖形之驗證，發現確如上述。並建議爾後所發展之危險評估量表均能清楚寫出再犯危險評估量表在預測2×2表上之ROC值及其敏感度與特異度，俾便量表之使用者及司法決策者能對量表之準確度有進一步的正確了解。

The risk assessment scale is usually used to predict the recidivism. It was always used the Pearson's r and Receiver Operative Characteristic Curve (ROC) to assess the predictive accuracy for the risk assessment scale. But what is the ROC, and what it means? How does the ROC relate to sensitivity and specificity? All the above were the topic of this study. During the development of risk assessment scale in Taiwan, it was always found that the ROC seemingly is the mean of sensitivity and specificity, while using 2 by 2 table of "guess yes or not" and "reoffense or not". In this study, it was verified the previous notion through algebra. The authors suggested that it has to mark the values of ROC, sensitivity, and specificity under the 2 by 2 predictive table in the future risk assessment scale to make forensic practitioners and criminal justice practitioners have a better understanding on the hit rates and false rates of the risk assessment scale.