【摘要】： 心理測驗中有許多干擾因素會影響測驗分數,影響考試的有效性和公平性。作弊是一個比較常見的因素,而在眾多的作弊方法中,作答抄襲是較難偵查的一種。作答抄襲指的是一個作答者從另一個作答者那里得到部分或者全部答案。為了偵查作答抄襲,研究者提出了很多方法。作答抄襲統計量是一種主要的偵查作答抄襲的方法。作答抄襲統計量應用于偵查作答抄襲可以追溯到上個世紀二十年代,隨后,研究者們提出了大量作答抄襲統計抄襲量指標。 本文介紹了早期的幾種主要的用于偵查作答抄襲的統計量指標:K,g,ω。在此基礎上,詳細地介紹了兩種偵查作答抄襲的新統計量指標:Kappa抄襲統計量指標和基于變換二項式的統計檢驗量指標的原理。本文采用稱名反應模型進行數據模擬,考察了Kappa抄襲統計量和基于變換二項式的統計檢驗量這兩種指標在測驗長度、作答者人數、項目抄襲率三個因素的不同水平上的偵查效果,比較兩種指標偵查作答抄襲時偵查率和Ⅰ型錯誤率的異同。 研究表明:(1)懷疑的抄襲者c抄襲懷疑的被抄襲者s的項目達到一定數量時,兩種指標對于作答抄襲的偵查率都非常高。抄襲率為20%時,Kappa抄襲統計量指標的偵查率已高達0.9;抄襲率為30%時,基于變換二項式的統計檢驗量指標的偵查率也高達0.9。(2)測驗長度,作答者樣本量,項目抄襲率對于兩種指標的偵查率都有相同的影響趨勢。測驗長度越長,項目抄襲率越高,偵查率越高。而作答者樣本量對偵查率則無明顯影響。(3)Kappa抄襲統計量的高偵查率以高Ⅰ型錯誤率為代價。錯誤偵查率為理論Ⅰ型錯誤率的3～5倍。基于變換二項式的統計檢驗量的錯誤偵查率則控制良好,均在理論Ⅰ型錯誤率值左右。
[Abstract]:There are many interference factors in psychological test that will affect the test score and the effectiveness and fairness of the test. Cheating is a more common factor, and among many cheating methods, plagiarism is difficult to detect. Plagiarism refers to the fact that one respondent gets part or all of the answer from another. In order to investigate plagiarism, the researchers put forward a number of methods. Plagiarism statistics are a major method of investigation and plagiarism. The application of plagiarism statistics in investigative plagiarism can be traced back to the 1920s, and then researchers put forward a large number of statistical plagiarism indicators. This paper introduces several main statistical indexes used in early investigation to answer plagiarism: K, g, 蠅. On this basis, this paper introduces in detail the principles of two new statistical indexes for investigation plagiarism: Kappa plagiarism statistics index and statistical test index based on transformation binomial. In this paper, the named reaction model is used to simulate the data, and the detection effects of Kappa plagiarism statistics and statistical tests based on transformation binomial on the test length, the number of respondents and the item plagiarism rate are investigated, and the similarities and differences between the detection rate and the type I error rate of the two indexes are compared. The results show that: (1) when the suspected plagiarist c plagiarizes the suspected plagiarist s, the detection rate of the two indicators for plagiarism is very high when the number of suspected plagiarism s items reaches a certain number. When the plagiarism rate is 20%, the detection rate of Kappa plagiarism statistics index is as high as 0.9, and when the plagiarism rate is 30%, the detection rate of statistical test index based on transformed binomial is also as high as 0.9. (2) the test length, the sample size of the respondent and the item plagiarism rate have the same influence trend on the detection rate of the two indexes. The longer the test length, the higher the plagiarism rate and the higher the detection rate. However, the sample size of respondents had no significant effect on the detection rate. (3) the high detection rate of Kappa plagiarism statistics was at the expense of high type I error rate. The error detection rate is 3 鈮，

本文編號：2507000