Ward L C, Ward J W, Moore C W
J Pers Assess. 1983 Aug;47(4):359-63. doi: 10.1207/s15327752jpa4704_4.
Multiple regression equations for estimating Wiggins content scores from 399- and 168-item abbreviations of Form R of the Minnesota Multiphasic Personality Inventory (MMPI) were derived for 100 patients and cross-validated in a separate sample of 50 patients. Across samples, Rs between 399-item short form scales and corresponding full scales were very high (median = .98 for the total sample), and the equations predicted accurately in cross-validation. Rs in estimating from 168 items were lower (median = .91 for the total sample), but the equations estimated mean content scores well in cross-validation. These results were viewed as indicating that there may be little loss in estimating from 399 items but that error in prediction may be unacceptably high for some of the 168-item short form scales. The correspondence of Wiggins scales and standard MMPI scales was also studied through regression analyses, and it was concluded that, although there is considerably overlap in the two sets of scales, inferences as to self-report content from the standard scales are not always simple or especially accurate.
从明尼苏达多相人格调查表(MMPI)R型的399项和168项缩写形式来估计威金斯内容分数的多元回归方程是针对100名患者得出的,并在另外50名患者的样本中进行了交叉验证。在各个样本中,399项简式量表与相应全量表之间的相关系数(Rs)非常高(总样本的中位数为0.98),并且这些方程在交叉验证中预测准确。从168项进行估计时的相关系数较低(总样本的中位数为0.91),但这些方程在交叉验证中对平均内容分数估计良好。这些结果被视为表明从399项进行估计可能损失不大,但对于某些168项简式量表而言,预测误差可能高得令人无法接受。还通过回归分析研究了威金斯量表与标准MMPI量表的对应关系,得出的结论是,尽管两组量表有相当大的重叠,但从标准量表推断自我报告内容并不总是简单或特别准确。