手机依赖的计算机化自适应测量及其效果评估

本文深入探讨了手机依赖的新的测量方式即计算机化自适应测量,并同时与原手机依赖的纸笔测验量表进行了比较,以探讨在相同测量长度下,新测量方式能在多大程度上提升对手机依赖的测量精度及测量信度。为此本文共进行了两项研究:研究1开发构建了手机依赖的计算机化自适应测量(CAT-MPD),并探讨了其测量的特征与性能。研究2通过对真实数据的模拟CAT测试,以CAT-MPD来源的纸笔测验量表为比较对象,考察CAT-MPD对原纸笔量表测量精度及测量信度的提升幅度。研究结果表明:CAT-MPD具有较理想的测量精度和信度,而且能有效减少被试所测项目量;同时,在同等条件下,CAT-MPD不论是在测量的精度还是测量的信度,均大幅优于手机依赖的纸笔测验量表。总之,本文为手机依赖的实际测量提供了一种新的技术支持。

Computerized Adaptive Measurement for Mobile Phone Dependence

This article explores in depth the new measurement method that mobile phone dependence, namely computerized adaptive testing for mobile phone dependence(CAT-MPD), and compares it with the paper-and-pencil test scales that assess mobile phone dependence to explore the extent to which the new measurement method can be used under the same measurement length. The proposed CAT-MPD aims to improve measurement accuracy and measurement reliability of mobile phone dependence. Under the framework of item response theory(IRT), this study firstly developed the item bank of mobile phone dependence for CAT-MPD and explored the characteristics and performance of its measurements in Study Ⅰ. We conducted sequential analyses to develop a high-quality item bank, which included unidimensionality test, IRT model selection, local independence test, item model fit test, item calibration and differential item functioning test. Finally, 90 items were included in the final item bank for CAT-MPD, in which all items had acceptable psychometric properties. Findings also indicated that under the termination rule of SE=.4, the correlation between the estimated trait value using 9.94 items and the estimated trait value estimated using 90 items of the entire project library was as high as.9(p<.001), and the measurement reliability of CAT-MPD is also as high as.87. That is to say, the proposed CAT-MPD has an acceptable reliability and can save a lot of items to decrease test burden if the termination rule set SE=.4, which is the termination rule we recommend. If you want a higher marginal reliability, you can use the termination rule of SE=.3. The correlation between the estimated trait value of 17.56 items and the estimated trait value of 90 items using the entire project library is.94(p <.001) and the marginal reliability reached.92. Due to the exposure control algorithm, the exposure rates of items are all below.3. All the items in the item bank have been used, that is, there are no items questions in the question bank. This shows that both the uniformity of the item usage and the safety of the item bank are ideal.In Study Ⅱ, we compared the CAT-MPD with the paper-and-pencil test scales of mobile phone dependence to explore whether the proposed CATMPD could greatly improve the measurement accuracy and measurement reliability, and significantly decrease the error of measurement when they use the same test length. The results show that the CAT-MPD 1) has a significant improvement on measurement accuracy and reliability(p<.001, Cohen’s d > 1) than the paper-and-pencil test scales of mobile phone dependence, 2) can effectively reduce the amount of items tested(reduced by 15.8%~31.6%, with an average of 24.4%), and 3) can greatly increase the measurement reliability by 5.7%~8.1%(average of 6.74%). All in all, under the same conditions(i.e., with same test length), CAT-MPD is superior to the paper-and-pencil tests of mobile phone dependence not only in measurement accuracy but also in measurement reliability. In summary, this article provides a new technical support for the actual measurement of mobile phone dependence.