基于统计学变量筛选方法的心理测验题目的维度识别

近年来多维心理测验被广泛应用于各类评估,虽然编制测验时知道整个测验考察的潜在特质(或称为维度),但是测验题目具体考察的维度仍需确定。借助多维项目反应理论模型与广义线性模型的关系,使用LASSO和弹性网两种变量筛选方法,可解决测验题目的维度识别问题。模拟研究发现,LASSO方法比弹性网方法具有更好的维度识别效果,前者对不同类型的多维测验具有较高的维度识别准确率。

Item Dimension Identification of Psychological Tests based on Statistical Variable Selection Methods

Multidimensional psychological tests have been widely used to evaluate examinees'latent traits in all kinds of subject assessment .Although the possible latent traits or the so‐called dimensions of the tests can be known to some extent ,the dimensions probed by each item of the tests are still needed to identify for the application purpose .Based on multidimensional item response theory and the shrinkage estimation methods of statistical variable selection ,this research explored to statistically identify the item‐dimension correspondence relationship in some typical psychological tests . Simulation studies were conducted to investigate the performance of the proposed method and the results showed that the method based on LASSO did better than that based on the elastic net in terms of correctly identifying the dimensions of test items .