本文選題：MPL模型　切入點(diǎn)：LASSO　出處：《數(shù)學(xué)的實(shí)踐與認(rèn)識》2017年21期 　論文類型：期刊論文

【摘要】：多維項目反應(yīng)理論(MIRT)模型是國際教育統(tǒng)計與心理測量學(xué)研究的熱點(diǎn)模型.在簡要介紹一種常見的MIRT模型和數(shù)理統(tǒng)計學(xué)熱門的變量篩選方法的基礎(chǔ)上,針對教育統(tǒng)計研究者常用的分?jǐn)?shù)減法測驗(yàn)數(shù)據(jù)進(jìn)行測驗(yàn)題目的維度識別.通過分別使用傳統(tǒng)的因子分析法、LASSO方法和彈性網(wǎng)方法分析測驗(yàn)數(shù)據(jù),獲得了測驗(yàn)題目的維度識別結(jié)果,并對它們的識別準(zhǔn)確率進(jìn)行比較.研究表明使用變量篩選方法尤其是LASSO方法能夠較好地識別該測驗(yàn)的題目維度間隸屬關(guān)系,為多維測驗(yàn)的維度識別提供可靠的信息.
[Abstract]:Multi-dimensional item response theory (RT) model is a hot research model in international educational statistics and psychometrics. Based on a brief introduction of a common MIRT model and a popular variable screening method in mathematical statistics, the paper presents a new method for selecting variables in international educational statistics and psychometrics. Based on the commonly used score subtraction test data of educational statistics researchers, the dimension recognition results of test questions are obtained by using the traditional factor analysis method and elastic net method, respectively, by using the traditional factor analysis method and the elastic network method to analyze the test data. The results show that the variable selection method, especially the LASSO method, can better identify the subordination relationship between the subject dimensions of the test, and provide reliable information for dimension recognition of multidimensional tests.
【作者單位】： 北京林業(yè)大學(xué)理學(xué)院;
【基金】：中央高校基本科研業(yè)務(wù)費(fèi)專項資金(2015ZCQ-LY-01) 國家自然科學(xué)基金青年科學(xué)基金項目(11701029);國家自然科學(xué)基金數(shù)學(xué)天元基金(11626040)
【分類號】：G40-051
，

本文編號：1577651