Dwyer A J
Department of Radiology, National Institutes of Health, Bethesda, MD 20814.
Radiology. 1991 Feb;178(2):328-30. doi: 10.1148/radiology.178.2.1987587.
Comparative studies of radiologic techniques commonly yield matched data due to the ease and desirability of performing all of the techniques on each of the patients. The two-by-two matched data array and the McNemar analysis provide a succinct format for the presentation and proper analysis of matched comparisons of binary (positive and negative) test results. When comparing tests, it is essential not to rely on just the statistical significance of the differences in sensitivities (or specificities); the magnitude of the differences must also be assessed. Confidence intervals provide a useful form of estimation by providing a range of plausible values and an indication of the precision of the estimate. The matched data array also indicates the complementarity of the diagnostic tests being compared.
由于对每位患者进行所有技术操作既简便又可取，放射学技术的比较研究通常会产生匹配数据。二乘二匹配数据阵列和麦克尼马尔分析为二元（阳性和阴性）检测结果的匹配比较提供了简洁的呈现和恰当分析格式。在比较检测时，至关重要的是不能仅仅依赖敏感性（或特异性）差异的统计学显著性；差异的大小也必须进行评估。置信区间通过提供一系列合理值和估计精度的指示，提供了一种有用的估计形式。匹配数据阵列还表明了所比较诊断检测的互补性。
