k系列指数、g_2、错同率的抄袭识别效能比较研究  

Comparision of Series of k- index,g_2- index,and False Same Rate of Answer Copying

作　　者：甘媛源[1,2] 余嘉元[2] 

机构地区：[1]江苏省教育评估院,南京210024 [2]南京师范大学心理学系,南京210097

出　　处：《心理学探新》2015年第5期464-470,共7页Psychological Exploration

基　　金：国家社会科学基金"十一五"规划课题(BBA080050)

摘　　要：采用蒙特卡洛模拟的方法进行k系列指数、g2、错同率的抄袭识别效能比较实验,探讨影响五种抄袭统计量抄袭识别效能的因素。结果表明:(1)在被试样本量、项目数、被抄者的能力水平、抄袭率四个因素中,抄袭率对抄袭识别效能的影响最大,被抄者的能力水平、项目数次之,被试样本量的影响最小;(2)在上述因素相当的情况下,g2的抄袭识别经验I型错误率远高于α水平,抄袭识别率也最高,k1、k2的经验I型错误率远低于α水平,抄袭识别率较高,错同率的经验I型错误率略高于α水平,抄袭识别率较低,k*'的经验I型错误率几乎均为0,抄袭识别率最低;(3)在被怀疑的被抄者已确定的情况下,k1、k2的抄袭识别效能最高,是较理想的抄袭统计量;(4)在同等条件下,当被怀疑的被抄者和抄袭者均未确定时,五种抄袭统计量的抄袭识别效能较被怀疑的被抄者已确定的情形均大幅下降;(5)在缺乏被怀疑者信息的情况下,需设定更为保守的指标临界值,以降低误判风险。In order to compare the power of answer copying detection of the answer copying statistics such as series of k - index ,g2 - in- dex, and false same rate, the comparison experiment was carried out under various copying conditions, sample size, test length, source＇ s ability, and copier＇ s detection rate. Which one was the most influencing factor of their answer copying detection power and which one was the best answer - copying statistics were investigated. The results indicated the following： （ 1 ） among sample size, test length, source＇ s ability, and copier＇ s detection rate ,the copier＇ s detection rate was the most influential factor of their answer - copying detec- tion power,test length and source＇s ability were the second ones,and sample size was the least one. （2）With the same sample size,test length, source＇ s ability, and copier＇ s detection rate, the empirical type I error rate of g2 - index was above the nominal ct level, and the detection rate of it was the highest ;k~ - index and k2 - index were able to hold the empirical type I error rates below the nominal ct lev- el,and their detection rates were the second higher;the empirical type I error rate of k＂ ＇ -index was close to O, and its detection rate was the lowest ;the empirical type I error rate of false same rate was slightly above the nominal ot level, and its detection rate was the second lower. （3）k1 -index and k2 -index were the best answer- copying statistics when the source under suspicion was determined, because their powers of answer copying detection were the highest. （4） With the same sample size, test length, source＇ s ability, and coplet＇s detection rate, the powers of answer copying detection of series of k -index,g2 -index, and false same rate were sharp decreased when the source and copier under suspicion were undetermined. （5）In order to reduce misjudgment risk, the conservative critical value of the answer - copying statistics would be determined when the examinees under suspicion w

关 键 词：k系列指数 G2 错同率 经验I型错误率 抄袭识别率 

分 类 号：B841.2[哲学宗教—基础心理学]