1、项目反应理论简介华东师范大学心理系文 剑 冰经典测量理论(CTT) 经典测量理论的假设 XTE 经典测量理论的信度 经典测量理论的效度 经典测量理论的试题参数 经典测量理论的测验编制经典测量理论的假设 观察分数真分数误差分数 XTE 观察分数与误差分数之间互相独立 误差分数的平均数为0 多次测量的误差分数之间相关为0经典测量理论的信度 信度的概念“真实分数方差在观测分数方差中所占的比率” 信度系数的估计方法重测信度(稳定性系数)复本信度(等值性系数)内部一致性信度评分者信度 信度系数的应用XXterSS1经典测量理论的试题参数 难度指标(通过率或得分率P值) 区分度指标(鉴别力指数D或相关系数
2、r) D PHPLmaxXXP 经典测量理论的测验编制 假设被试的特质是正态分布,从而测验总分的分布也是正态 测验分数尽可能区分被试,因此测验总分的变异程度越大越好 测验中试题的难度中等为好,区分度越大越好经典测量理论的缺陷 参数依赖于样本能力量表与难度量表不统一对于所有被试的测量误差相等无法反应潜在特质与被试作答之间的关系在测验编制问题上的困惑准备知识 标准分数 Z0,高于平均,Z0,低于平均 P(-1.96Z1.96)=0.950 P(-3ZCOMMENTGLOBAL DFN=C:YAN2.DAT, NIDW=5, NPARM=2, SAVE;SAVE SCO = YAN2.SCO, PA
3、RM = YAN2.PAR, TST = YAN2.TST,IST=YAN2.IST;LENGTH NITEMS=(80);INPUT NTOT=80,NALT=4, KFN=KEY.TXT,OFN=OMIT.TXT;(5A1,80A1)CALIB NQPT=40, CYC=100, NEW=30, CRIT=.001, PLOT=0;SCORE MET=2, IDIST=0, RSC=0, INF=1;Title lineBILOG 程序文件 (*.BLM)IRT calibration of chinese and maths score.COMMENTGLOBAL DFN=C:YAN2.
4、DAT, NIDW=5, NPARM=2, SAVE;SAVE SCO = YAN2.SCO, PARM = YAN2.PAR, TST = YAN2.TST,IST=YAN2.IST;LENGTH NITEMS=(80);INPUT NTOT=80,NALT=4, KFN=KEY.TXT,OFN=OMIT.TXT;(5A1,80A1)CALIB NQPT=40, CYC=100, NEW=30, CRIT=.001, PLOT=0;SCORE MET=2, IDIST=0, RSC=0, INF=1;数据文件名个人ID位数模型参数个数保存外部文件BILOG 程序文件 (*.BLM)IRT c
5、alibration of chinese and maths score.COMMENTGLOBAL DFN=C:YAN2.DAT, NIDW=5, NPARM=2, SAVE;SAVE SCO = YAN2.SCO, PARM = YAN2.PAR, TST = YAN2.TST,IST=YAN2.IST;LENGTH NITEMS=(80);INPUT NTOT=80,NALT=4, KFN=KEY.TXT,OFN=OMIT.TXT;(5A1,80A1)CALIB NQPT=40, CYC=100, NEW=30, CRIT=.001, PLOT=0;SCORE MET=2, IDIST
6、=0, RSC=0, INF=1;保存试题参数,被试参数,CTT结果,测验信息函数BILOG 程序文件 (*.BLM)IRT calibration of chinese and maths score.COMMENTGLOBAL DFN=C:YAN2.DAT, NIDW=5, NPARM=2, SAVE;SAVE SCO = YAN2.SCO, PARM = YAN2.PAR, TST = YAN2.TST,IST=YAN2.IST;LENGTH NITEMS=(80);INPUT NTOT=80,NALT=4, KFN=KEY.TXT,OFN=OMIT.TXT;(5A1,80A1)CALI
7、B NQPT=40, CYC=100, NEW=30, CRIT=.001, PLOT=0;SCORE MET=2, IDIST=0, RSC=0, INF=1;(分)测验题数BILOG 程序文件 (*.BLM)IRT calibration of chinese and maths score.COMMENTGLOBAL DFN=C:YAN2.DAT, NIDW=5, NPARM=2, SAVE;SAVE SCO = YAN2.SCO, PARM = YAN2.PAR, TST = YAN2.TST,IST=YAN2.IST;LENGTH NITEMS=(80);INPUT NTOT=80,
8、NALT=4, KFN=KEY.TXT,OFN=OMIT.TXT;(5A1,80A1)CALIB NQPT=40, CYC=100, NEW=30, CRIT=.001, PLOT=0;SCORE MET=2, IDIST=0, RSC=0, INF=1;omit文件名总题数选项个数标准答案文件名BILOG 程序文件 (*.BLM)IRT calibration of chinese and maths score.COMMENTGLOBAL DFN=C:YAN2.DAT, NIDW=5, NPARM=2, SAVE;SAVE SCO = YAN2.SCO, PARM = YAN2.PAR,
9、TST = YAN2.TST,IST=YAN2.IST;LENGTH NITEMS=(80);INPUT NTOT=80,NALT=4, KFN=KEY.TXT,OFN=OMIT.TXT;(5A1,80A1)CALIB NQPT=40, CYC=100, NEW=30, CRIT=.001, PLOT=0;SCORE MET=2, IDIST=0, RSC=0, INF=1;FORTRAN 语言读数据的格式A,X,T,I,/BILOG 程序文件 (*.BLM)IRT calibration of chinese and maths score.COMMENTGLOBAL DFN=C:YAN2.
10、DAT, NIDW=5, NPARM=2, SAVE;SAVE SCO = YAN2.SCO, PARM = YAN2.PAR, TST = YAN2.TST,IST=YAN2.IST;LENGTH NITEMS=(80);INPUT NTOT=80,NALT=4, KFN=KEY.TXT,OFN=OMIT.TXT;(5A1,80A1)CALIB NQPT=40, CYC=100, NEW=30, CRIT=.001, PLOT=0;SCORE MET=2, IDIST=0, RSC=0, INF=1;试题参数估计时的设定画出拟合度差(pCOMMENTGLOBAL DFN=C:YAN2.DAT
11、, NIDW=5, NPARM=2, SAVE;SAVE SCO = YAN2.SCO, PARM = YAN2.PAR, TST = YAN2.TST,IST=YAN2.IST;LENGTH NITEMS=(80);INPUT NTOT=80,NALT=4, KFN=KEY.TXT,OFN=OMIT.TXT;(5A1,80A1)CALIB NQPT=40, CYC=100, NEW=30, CRIT=.001, PLOT=0;SCORE MET=2, IDIST=0, RSC=0, INF=1;被试能力估计时的设定1-ML2-EAP(缺省)3-MAP0-不做重新标刻(缺省)1-按scale和
12、location线性变换3-按样本的L和S重新标刻3-EAP时潜变量以L为均数S为标准差测验信息曲线BILOG 结果文件 (*.PH1) ITEM STATISTICS FOR SUBTEST TEST0001 ITEM*TEST CORRELATION ITEM NAME #TRIED #RIGHT PCT LOGIT PEARSON BISERIAL - 1 ITEM0001 480.0 395.0 82.3 -1.54 0.318 0.468 2 ITEM0002 480.0 357.0 74.4 -1.07 0.306 0.415 3 ITEM0003 480.0 444.0 92.5
13、 -2.51 0.252 0.469 4 ITEM0004 480.0 321.0 66.9 -0.70 0.468 0.608 5 ITEM0005 480.0 292.0 60.8 -0.44 0.119 0.151 6 ITEM0006 480.0 265.0 55.2 -0.21 0.162 0.204 7 ITEM0007 480.0 315.0 65.6 -0.65 0.288 0.372 8 ITEM0008 480.0 247.0 51.5 -0.06 0.391 0.490 9 ITEM0009 480.0 178.0 37.1 0.53 0.128 0.163 10 ITE
14、M0010 480.0 253.0 52.7 -0.11 0.406 0.509CTT的试题参数Ln(1-p)/pBILOG 结果文件 (*.PH2) CYCLE 15; LARGEST CHANGE= 0.00007 SUBTEST TEST0001; ITEM PARAMETERS AFTER CYCLE 15 ITEM INTERCEPT SLOPE THRESHOLD LOADING ASYMPTOTE CHISQ DF S.E. S.E. S.E. S.E. S.E. (PROB) - ITEM0001 | 1.785 | 0.922 | -1.936 | 0.678 | 0.000
15、 | 2.2 8.0 | 0.147* | 0.146* | 0.265* | 0.107* | 0.000* | (0.9758) | | | | | | ITEM0002 | 1.214 | 0.816 | -1.487 | 0.632 | 0.000 | 3.4 9.0 | 0.118* | 0.124* | 0.224* | 0.096* | 0.000* | (0.9469) | | | | | | IRT的试题参数-Slope*thresholdSlope/sqrt(1+slope2)BILOG 结果文件 (*.PH3) GROUP SUBJECT IDENTIFICATION M
16、ARGINAL WEIGHT TEST TRIED RIGHT PERCENT ABILITY S.E. PROB - 1 11 | | 1.00 TEST0001 80 46 57.50 | -0.4595 0.1175 | 0.00 1 12 | | 1.00 TEST0001 80 46 57.50 | -0.5095 0.2318 | 0.00 1 13 | | 1.00 TEST0001 80 28 35.00 | -1.7741 0.4445 | 0.00 1 14 | | 1.00 TEST0001 80 58 72.50 | -0.2157 0.3886 | 0.00 1 15
17、 | | 1.00 TEST0001 80 57 71.25 | 0.0378 0.4430 | 0.00 1 16 | | 1.00 TEST0001 80 20 25.00 | -2.2754 0.2127 | 0.00 1 17 | | 1.00 TEST0001 80 63 78.75 | 0.4364 0.1461 | 0.00 1 18 | | 1.00 TEST0001 80 65 81.25 | 0.5205 0.2539 | 0.00被试的能力参数试题参数文件 (*.PAR)BILOG保存的外部文件 试题参数文件(*.PAR) 被试能力估计文件(*.SCO)数据格式与PH2和PH3文件中基本相同