Levin Joel R, Ferron John M, Gafurov Boris S
a Department of Educational Psychology , University of Arizona , Tucson , AZ , USA.
b Educational Measurement and Research , University of South Florida , Tampa , FL , USA.
Dev Neurorehabil. 2018 Jul;21(5):290-311. doi: 10.1080/17518423.2016.1197708. Epub 2016 Jul 1.
In three simulation investigations, we examined the statistical properties of several different randomization-test procedures for analyzing the data from single-case multiple-baseline intervention studies. Two procedures (Wampold-Worsham and Revusky) are associated with single fixed intervention start points and three are associated with randomly determined intervention start points. Of the latter three, one (Koehler-Levin) is an existing procedure that has been previously examined and the other two (modified Revusky and restricted Marascuilo-Busk) are modifications and extensions of existing procedures. All five procedures were found to maintain their Type I error probabilities at acceptable levels. In most of the conditions investigated here, two of the random start-point procedures (Koehler-Levin and restricted Marascuilo-Busk) were more powerful than the others with respect to detecting immediate abrupt intervention effects. For designs in which it is not possible to include the same series lengths for all cases, either the modified Revusky or restricted Marascuilo-Busk procedure is recommended.
在三项模拟研究中，我们检验了几种不同随机化检验程序的统计特性，用于分析单病例多基线干预研究的数据。两种程序（万波德 - 沃舍姆和雷夫斯基）与单个固定干预起始点相关，三种程序与随机确定的干预起始点相关。在与随机确定干预起始点相关的三种程序中，一种（克勒 - 莱文）是先前已被检验过的现有程序，另外两种（改良雷夫斯基和受限马拉斯奎洛 - 布斯克）是对现有程序的修改和扩展。发现所有五种程序的I型错误概率均维持在可接受水平。在此处研究的大多数条件下，在检测即时突发干预效果方面，两种随机起始点程序（克勒 - 莱文和受限马拉斯奎洛 - 布斯克）比其他程序更具效力。对于无法为所有病例纳入相同序列长度的设计，建议采用改良雷夫斯基或受限马拉斯奎洛 - 布斯克程序。
