Development of an computerized adaptive testing of New Japanese Version of Minnesota Multiphasic Personality Inventory (MMPI).

自20世纪60年代以来，许多研究已经将明尼苏达多相人格调查表（MMPI）的管理计算机化。然而，现在，我们应该开发一个计算机化的MMPI自适应测试。本研究应用项目反应理论（IRT）编制了MMPI（新日文版）自适应量表，并利用IRT的双参数Logistic模型，对日本全国常模组（522名女性和500名男性）的项目难度和区分度参数进行了估计。除Mf和Si不用于临床决策外，表明每个量表在最大测试信息或最大区分力下的性状值接近2.0，相当于T得分的70。该图表示MMPI的异常，说明MMPI可以通过IRT区分异常组和正常组。本研究的下一步是使用与正常组不同的样本验证这些项目参数的交叉效度。一个真实的数据模拟证明了两个发现。首先，估计的高特质与真实的高量表得分很好地拟合。第二，自适应MMPI的管理项目的数量从所有减少到十分之一。因此，能够以较少的项目准确地判断异常，并且适合于筛选测试。

Many studies have been computerizing an administration of the Minnesota Multiphasic Personality Inventory (MMPI) from 1960s. However, now, we should develop a computerized adaptive testing of the MMPI. In this study, Item Response theory (IRT) was applied to develop the adaptive MMPI (New Japanese Version).Through the two-parameter-logistic model of IRT, item difficulty and discrimination parameters were estimated from Japanese national norm group data (522 females and 500 males). Except for the Mf and Si which are not utilized for clinical decision, it was indicated that a trait value at maximum test information or maximum discrimination power of each scale was nearly 2.0, which is equal to 70 on T-score. The figure means abnormality on the MMPI, so it is revealed that through IRT the MMPI can distinguish abnormal group from normal one.The next step of this study verified cross-validity of those item parameters using a distinct sample from the norm group. A real data simulation demonstrated two finds. First, an estimated high trait was good fitting to a real high scale score. Second, the adaptive MMPI reduced the number of administered items from all to one tenth. So, it is able to accurately judge abnormality with fewer items, and is suitable for a screening test.