作者: David L Streiner , John Cairney
DOI: 10.1177/070674370705200210
关键词:
摘要: It is often necessary to dichotomize a continuous scale separate respondents into normal and abnormal groups. However, because the distributions of scores in these 2 groups most overlap, any cut point that chosen will result types errors: false negatives (that is, cases judged be normal) positives placed group). Changing alter numbers erroneous judgments but not eliminate problem. A technique called receiver operating characteristic (ROC) curves allows us determine ability test discriminate between groups, choose optimal point, compare performance or more tests. We discuss how calculate ROC factors must considered choosing an point.