2015, Vol. 47, No.1, 129140 Acta Psychologca Snca DOI: 10.3724/SP.J.1041.2015.00129 CD-CAT * 1,2 3 2,4 ( 1, 400715) ( 2, 100875) ( 3,, 61820 ) ( 4, 100875),, 4 CD-CAT, (SEA)(DAPP)(HA) (HM), HSU KL (1),,, (2), 4, HSU,,, (3), 6,,, (4), SEAHM HA HSU, KL DAPP ; ; ; ; DINA B841 1, (Cogntve Dagnostc Computerzed Adaptve Testng, CD-CAT)CD-CAT CAT,,, (Knowledge State, KS),,, CD-CAT CAT, (1); (2) ; (3); (4); (5)(, 2011) CD-CAT,, CD-CAT (Cheng, 2009, 2010; Wang, 2013; Wang, Chang, & Douglas, 2012; Xu, Chang, & Douglas, 2003;,, 2011)(Wang, Chang, & Huebner, 2011;, 2011;,, 2013) (Chen, Xn, Wang, & Chang, 2012;,, 2011;,,, 2011), CD-CAT, Hsu, Wang Chen (2013) CD-CAT, (fxed-length) CAT,,, CAT,,, CAT, : 20131030 * (20120003110002) :, E-mal: banyufang66@126.com 129
130 47 (Wess & Kngsbury, 1984) CAT (varable-length)cat, CAT, (Babcock & Wess, 2009)Kngsbury Houser (1993), CAT 0-1 CAT, CAT CAT CAT, CAT,, CAT, CAT (Cho, Grady, & Dodd, 2010; Dodd, 1990; Dodd, Koch, & De Ayala, 1993; Dodd, De Ayala, & Koch, 1995),, (1), ; (2), (, 2011) CD-CAT, Hsu (2013) Tatsuoka (2002), CD-CAT ( HSU, 2 ),, ;, CD-CAT CAT, CD-CAT, Wang (2011), CD-CAT,,, CD-CAT,, CD-CAT, (,, 2013;,,, 2012) CD-CAT CAT, CAT, ( Q ),,, CD-CAT, Hsu (2013) CD- CAT, CD-CAT,, CD-CAT, Cheng (2008) KL, CAT, CD-CAT, (standard error of attrbute, SEA) (halvng algorthm, HA) (dfference of the adjacent posteror probablty method, DAPP), Tatsuoka (2002), (hybrd method, HM) ( 2 ) DINA (Junker & Sjtsma, 2001), ( 3 ), HSU Cheng KL, CD-CAT 2 CD-CAT 6, HSU, SEA HA ;, DAPP KL ;,, HA (tem-level), ; 5 (examnee-level), 6 2.1 HSU Tatsuoka (2002) CD-CAT, 0.8, Hsu (2013) Tatsuoka, CD-CAT, P 1st (, 0.7), P 2nd (, 0.1), 2.2 (standard error of attrbute method, SEA),, KS
1 CD-CAT 131, KS, KS, CDM (partally ordered classfcaton models; Tatsuoka, 2002), (partally ordered sets) (Zhang & Ip, 2012), KS, (approxmaton) KS,, (Structural Equaton Modelng, SEM), SEM, K KS, Pk,mn( Pk) k 1 ( Pk 1), KS, K k k 1 P,,,, P k (Rupp, Templn, & Henson, 2010; P242), KS SE( ) P (1 P ) (1) k k k, SE( k ) k, k 1, 2,, K P k SEA k,, (, 0.2), 2.3 (halvng algorthm, HA) Tatsuoka Ferguson (2003),, HA,,,,, t t,,, p t, t, ( c) ( c 1, 2,, 2 K ), qh c: cqh qhqh, HA HAh p, t(1 p, t) HA (, 0.1), 2.4 (dfference of the adjacent posteror probablty method, DAPP) CD-CAT,, (Cheng, 2009),, HSU P 1st DAPP, (P 1st ) t, 1 t P1st ( ) P 1st ( ) t (t t P 1 st( ) t, KS ), 2.5 KL Cheng (2008) KL KL, KL( t, t, 1) ( t, t ), 2.6 (hybrd method, HM) Hsu (2013), P 1st P 2nd ( Tatsuoka (2002)), HM P 1st, DAPP, t 1 t P1st ( ) P1st ( ) (t t t P1 st ( ) t, KS ), 3 3.1 DINA DINA, s g X j j ( Xj 1, Xj 0 ), j j 1 j j j gj PX ( 1 ) (1 s ) (2)
132 47 K, j k k 1 q jk,, j, j 1,, j 0 k k, k 1, k 0 q jk j k, q jk 1, q jk 0 sj P( Xj 0 j 1), j, ; g P( X 1 0) j j j, j, s j g j, s g (Templn & Henson, 2006) 1 j j 3.2, KL (PWKL), KL (HKL)(SHE) (Cheng, 2009;, 2011) Hsu (2013), PWKL CD-CAT PWKL K 2 1 t t h h h l l1 x0 ( t h ), t( l) PWKL ( ) log[ P( X x )/ P( X x )], PX x (3), h, x t 1 t t,, ( ) PWKL, l PWKL 3.3 Wang (2011) CAT, (mportance parameter), CD-CAT (Restrctve Progressve method, RP) (Restrctve Threshold method, RT)RP ( RT ) 1 x / L (x, L ) (), 1 x / L, 1,, ;,, t l,, L,, CD-CAT RP RT ( 3.3.2 3.3.3 ), (Modfed Restrctve Progressve, MRP) (Modfed Restrctve Threshold, MRT), CD-CAT (smple), CD-CAT, 3.3.1 smple smple PWKL f h (, 2011), f h rmax nh / N (4) r max, rmax, nh h, N 3.3.2 MRP CD-CAT, RP, MRP, Wang (2011), RP, f h MRP, f h, MRP MRP _ PWKLh MRP _ PWKLh (1exp h /r max )[( P1 st Pcurrent )/ P1 st R h PWKL P / P ], h S (5) h current 1st h, P 1st, Pcurrent S h, exp h h H * S h *, Rh ~ U(0, H ) 3.3.3 MRT ( PWKL ), ()
1 CD-CAT 133 [max( PWKL),max( PWKL)], [max( PWKL) mn( PWKL )] (1 P / P ), MRT, current 1st r max, P current,, Pcurrent P current, MRT,, 4 Matlab (R2011b), 4.1 (2011)(2011), Q, Q 1 Q 2 Q 3, 6, 360 U (0.05, 0.25) g j s j 2000 0.5 4.2 KS DINA j P j, U (0,1) m P j m, j 1, 0 CD-CAT,,, (Maxmum A Posteror, MAP) 4.3 (1)P 1st 0.8 0.9; P 2nd 0.002 0.003; (2),, (,, ),, HM DAPP KL, 4 0.050.010.005 0.001; SEA, 5 0.30.250.20.1 0.05; HA, 5 0.1 0.050.010.005 0.001; (3), r max 0.2, 2, CD-CAT,, 30, 4 (smple MRP MRT ) 6,, HSU 4 SEA 5 HA 5 DAPP 4 KL 4 HM 4, 30 4 (4 5 5 4 4 4) 30 3120 4.4 (Pattern Correct Classfcaton Rate, PCCR), N PCCR t / N (6) 1 PCCR ( ( 1, 2,, K ) ) K, N, () X, Z, X Z, t 1; t 0, (,, ) 5 5.1 1, 6 CD-CAT 1, 6,, P 1st,,, PCCR,,,, HSU, P 1st PCCR P 1st P 2nd PCCR (0.8394< 0.9968, 0.9219<0.9980),,,, Hsu (2013) HM,
134 47 1 6 (30 ) TR P1 P2 epslon M SD MAX MIN # NU PCCR HSU 0.8 9.2 2.6 27 5 228 0.8394 0.8 0.003 18.5 5.1 43 8 158 0.9968 0.9 11.4 3.4 32 6 214 0.9219 0.9 0.002 18.3 5.1 52 8 141 0.9980 HM 0.8 0.05 12.0 3.4 31 7 223 0.9357 0.8 0.01 14.6 3.7 37 8 187 0.9699 0.8 0.005 15.5 3.8 36 8 183 0.9713 0.8 0.001 19.5 4.2 37 8 179 0.9914 0.9 0.05 12.9 3.3 32 7 195 0.9586 0.9 0.01 14.8 3.7 36 8 171 0.9791 0.9 0.005 16.6 4.1 35 8 167 0.9893 0.9 0.001 18.4 4.6 41 9 157 0.9911 SEA 0.3 8.6 2.4 27 5 262 0.7963 0.25 10.9 2.7 27 5 234 0.8988 0.2 12.7 3.6 32 6 216 0.9672 0.1 14.2 3.8 32 6 174 0.9796 0.05 18.0 4.9 44 7 165 0.9927 HA 0.1 8.7 2.4 23 5 238 0.8084 0.05 12.3 3.2 38 6 203 0.9403 0.01 14.0 4.3 48 7 163 0.9816 0.005 15.9 4.2 42 8 148 0.9913 0.001 20.25 5.3 53 10 108 0.9982 DAPP 0.05 5.6 5.0 34 2 240 0.3387 0.01 15.9 4.4 42 8 156 0.9885 0.005 18.3 5.2 45 9 138 0.9922 0.001 21.1 5.8 47 10 102 0.9989 KL 0.05 11.3 3.1 32 7 205 0.9052 0.01 14.8 3.6 41 8 173 0.9752 0.005 15.9 3.9 40 9 146 0.9801 0.001 21.2 5.2 53 10 118 0.9957 TR (Termnaton Rules), P1, P2, epslon, M, SD, MAX, MIN, # NU =0.001 HSU ( 1 8 2, 12 4 ); =0.05, 2.8 (12.0 9.2) 1.5 (12.9 11.4), PCCR 9.63% (0.9357 0.8394) 3.67% (0.9586 0.9219) SEA, =0.3, PCCR 0.7963, 8.6 ; =0.25, PCCR 0.9, =0.2, PCCR 0.9672, HSU, 1.3 (12.7 11.4), PCCR 4.53%; =0.05, PCCR 0.9927, HSU DAPP, =0.05, PCCR, 0.3387, 5.6 ; =0.01, PCCR 0.9885, 15.9, 0.001, PCCR 0.9989, 240 102 KL HA DAPP (SEA),, 4 KL HSU, (HA ), (HSU, DAPP, HM, SEA KL ), CD-CAT 5.2 2 4 smplemrt MRP, 6,,, PCCR (p) PCCR (max), Hsu (2013) P 1st,, (, MRT MRP 0), PCCR (p), (%max),, %max,, SEA smple %max 14.9 ( =0.05 ), MRT MRP %max 65.85 45.60, PCCR (p), 0.9951, 0.9971 0.9975HA smple %max, MRP, MRT, PCCR (p) 1, KL MRP %max, smple, %max MRT, PCCR (p) smple MRT 0.98, MRP PCCR (p) 0.7802, smple ( 2), DAPP =0.05, PCCR (p) 0.3361, 6 MRT ( 3),
1 CD-CAT 135 2 smple 6 (30 ) TR P1 P2 epslon M Max (r) T # NU %max PCCR (p) PCCR (max) HSU 0.8 11.6 0.1135 0.0665 124 0.2 0.8558 0.7867 0.8 0.003 21.5 0.1315 0.0855 38 11.5 0.9942 0.9191 0.9 14.1 0.1180 0.0700 80 0.6 0.9349 0.8326 0.9 0.002 22.3 0.1315 0.0875 31 12.4 0.9979 0.9204 HM 0.8 0.05 15.5 0.1275 0.0803 79 1.14 0.9426 0.7653 0.8 0.01 16.9 0.1250 0.0746 59 2.95 0.9538 0.8916 0.8 0.005 18.4 0.1275 0.0788 55 3.15 0.9536 0.8642 0.8 0.001 19.6 0.1280 0.0810 47 5.95 0.9932 0.9322 0.9 0.05 15.9 0.1240 0.0729 69 1.35 0.9669 0.7674 0.9 0.01 18.5 0.1250 0.0798 57 4.8 0.9773 0.8479 0.9 0.005 20.0 0.1295 0.0807 41 6.65 0.9761 0.9191 0.9 0.001 22.6 0.1345 0.0864 43 10.85 0.9967 0.9335 SEA 0.3 11.5 0.1205 0.0645 119 0.15 0.8398 0.7333 0.25 13.2 0.1205 0.0707 96 0.45 0.9219 0.6411 0.2 14.1 0.1150 0.0712 82 0.7 0.9532 0.7522 0.1 18.0 0.1285 0.0797 60 4.7 0.9825 0.8910 0.05 20.5 0.1365 0.0867 29 14.9 0.9951 0.9383 HA 0.1 10.3 0.1135 0.0640 127 0.2 0.8014 0.7500 0.05 14.1 0.1205 0.0726 84 0.7 0.9455 0.8569 0.01 17.7 0.1270 0.0806 50 4.55 0.9832 0.8476 0.005 19.8 0.1335 0.0854 41 9.75 0.9929 0.9114 0.001 21.7 0.1345 0.0894 25 24.0 0.9980 0.9580 DAPP 0.05 5.7 0.0925 0.0511 183 0.25 0.3361 0.8333 0.01 18.8 0.1280 0.0811 49 6.75 0.9873 0.8802 0.005 20.1 0.1275 0.0817 33 12.75 0.9939 0.9320 0.001 23.5 0.1365 0.0903 25 27.95 0.9987 0.9590 KL 0.05 14.6 0.1385 0.0786 79 0.15 0.9085 0.7711 0.01 18.1 0.1250 0.0818 46 4.6 0.9726 0.8778 0.005 20.6 0.1305 0.0838 45 9.1 0.9945 0.9175 0.001 24.2 0.1385 0.0883 22 26.2 0.9974 0.9474 Max (r), T, %max, PCCR (p), PCCR (max) DAPP, = 0.005, PCCR (p) 0.6438, =0.001, PCCR (p) 0.9823, %max 47%, DAPP MRT KL,, =0.05, PCCR (p) 0.3658, 0.01 0.001, PCCR (p) 0.8395 0.9873 SEAHA KL,, PCCR (p), %max,,, ;, MRT (overcontrol), Max (r) r max = 0.2,, MRP ( 4), DAPP, =0.001, PCCR (p) 0.6724, KL, PCCR (p) 0.7802, MRP,, %max MRT, PCCR (p) MRT, Wang (2011) CD-CAT,, 2 4, PCCR (max) PCCR (p)
136 47 3 MRT 6 (30 ) TR P1 P2 epslon M Max (r) T # NU %max PCCR (p) PCCR (max) HSU 0.8 12.6 0.1995 0.0728 0 0.2 0.8637 0.6953 0.8 0.003 20.3 0.1995 0.1207 0 8.15 0.9950 0.9472 0.9 15.0 0.1995 0.0853 0 0.55 0.9368 0.7119 0.9 0.002 21.3 0.1995 0.1207 0 11.25 0.9965 0.9609 HM 0.8 0.05 16.0 0.1995 0.0969 0 0.55 0.9555 0.8660 0.8 0.01 18.5 0.1995 0.1116 0 3.00 0.9626 0.9000 0.8 0.005 19.0 0.1995 0.1145 0 4.45 0.9631 0.9122 0.8 0.001 21.4 0.1995 0.1200 0 8.4 0.9942 0.9454 0.9 0.05 16.5 0.1995 0.0914 0 1.35 0.9597 0.8320 0.9 0.01 19.3 0.1995 0.1092 0 6.55 0.9838 0.9045 0.9 0.005 20.1 0.1995 0.1114 0 8.2 0.9830 0.9310 0.9 0.001 23.2 0.1995 0.1217 0 13.1 0.9939 0.9385 SEA 0.3 17.3 0.0975 0.0636 0 5.75 0.9026 0.7335 0.25 20.8 0.1065 0.0685 0 10.45 0.9353 0.7598 0.2 21.1 0.1120 0.0720 0 24.9 0.9555 0.7897 0.1 25.3 0.1270 0.0913 0 35.6 0.9867 0.8629 0.05 27.1 0.1395 0.0863 0 65.85 0.9971 0.9080 HA 0.1 20.4 0.0960 0.0628 0 12.85 0.9154 0.7021 0.05 21.1 0.1160 0.0712 0 21.55 0.9476 0.8001 0.01 23.2 0.1240 0.0817 0 45.9 0.9883 0.8617 0.005 24.2 0.1325 0.0848 0 52.05 0.9947 0.8736 0.001 26.2 0.1465 0.0895 0 71.4 0.9995 0.8855 DAPP 0.05 2.7 0.0135 0.0078 0 0.1220 0.01 8.3 0.0395 0.0288 0 1.55 0.3430 0.6251 0.005 12.1 0.0570 0.0355 0 8.95 0.6438 0.8300 0.001 25.9 0.1305 0.0824 0 46.75 0.9823 0.9028 KL 0.05 9.0 0.0505 0.0299 0 0.05 0.3658 0.5833 0.01 19.9 0.0960 0.0642 0 6.5 0.8395 0.7612 0.005 21.7 0.1160 0.0732 0 14.2 0.9218 0.8072 0.001 27.0 0.1285 0.0931 0 50.65 0.9873 0.8834 2,, DAPP KL,,, (, P 1st 0.8), P 1st 0.8 (, =0.01), P 1st, 1 2 DAPP ( =0.05),, KS 2 K, 2 K DAPP, A 18, P 1st, 0.95, KS, A KS B 4, P 1st, P 1st ( 0.07 ), KS,, B KS 3 4 DAPP ( =0.05),, KS C 24,, KS 25, 30, 0.9, KS,, C KS D 30, P 1st 0.5
1 CD-CAT 137 4 MRP 6 (30 ) TR P1 P2 epslon M Max (r) T # NU %max PCCR (p) PCCR (max) HSU 0.8 14.3 0.0925 0.0444 0 1.3 0.8705 0.6339 0.8 0.003 23.1 0.1215 0.0690 0 20.2 0.9950 0.9082 0.9 16.3 0.1030 0.0504 0 3.1 0.9352 0.7878 0.9 0.002 24.1 0.1250 0.0718 0 27.4 0.9985 0.9024 HM 0.8 0.05 17.6 0.1035 0.0445 0 4.15 0.9497 0.7731 0.8 0.01 19.8 0.1150 0.0632 0 8.05 0.9562 0.8961 0.8 0.005 20.7 0.1115 0.0645 0 11.85 0.9594 0.8833 0.8 0.001 24.5 0.1155 0.0687 0 19.45 0.9949 0.9192 0.9 0.05 18.0 0.0985 0.0539 0 3.45 0.9531 0.7893 0.9 0.01 21.0 0.1165 0.0658 0 15.6 0.9824 0.8654 0.9 0.005 21.2 0.1230 0.0688 0 20.2 0.9811 0.8964 0.9 0.001 23.8 0.1275 0.0714 0 26.95 0.9913 0.9270 SEA 0.3 16.3 0.0535 0.0544 0 3.8 0.8428 0.6701 0.25 18.2 0.0565 0.0636 0 6.45 0.8938 0.7365 0.2 20.7 0.0625 0.0688 0 8.35 0.9435 0.7611 0.1 23.9 0.0820 0.0786 0 27.55 0.9881 0.8687 0.05 25.6 0.0770 0.0736 0 45.6 0.9975 0.8701 HA 0.1 17.1 0.0555 0.0517 0 7.2 0.8714 0.6676 0.05 18.8 0.0620 0.0588 0 15.6 0.9469 0.7316 0.01 21.2 0.0720 0.0691 0 39.5 0.9809 0.8243 0.005 22.1 0.0745 0.0717 0 46.4 0.9936 0.8629 0.001 23.7 0.0800 0.0771 0 68.45 0.9994 0.8640 DAPP 0.05 3.4 0.0135 0.0090 0 0.05 0.1712 0.6164 0.01 9.1 0.0290 0.0251 0 0.1 0.3880 0.7894 0.005 12.1 0.0365 0.0331 0 1.15 0.5248 0.8821 0.001 18.0 0.0550 0.0517 0 9.4 0.6724 0.9210 KL 0.05 8.6 0.0275 0.0235 0 0.05 0.4227 0.6867 0.01 13.6 0.0405 0.0372 0 1.65 0.6171 0.8206 0.005 15.5 0.0465 0.0429 0 2.25 0.6890 0.8525 0.001 18.9 0.0585 0.0550 0 9.6 0.7802 0.8854 2 1 A 2 B
138 47 3 C 4 D, KS, KS, 6 CD-CAT CAT,,,,,, CD-CAT, CD-CAT,, CD-CAT CAT (varablelength) CAT, CAT, (Babcock & Wess, 2009) 4 CD-CAT SEA HA DAPP HM, HSU KL 4 CD-CAT, (1)6,,,, (2), 4, HSU,,,,, (3), 6, MRT MRP, 0,,,, (4), SEAHM HA HSU, KL DAPP,, SEA,, KS, KS, KS,, CD-CAT, g s U (0.05,0.25),, g s (de la Torre, 2009;, 2012), 6,, 6, (Leghton, Gerl, & Hunka,
1 CD-CAT 139 2004),,, 6,,, 6,,, ; ( 30 ),, (Mao & Xn, 2013) CD-CAT Babcock, B., & Wess, D. J. (2009). Termnaton crtera n computerzed adaptve tests: Varable-length cats are not based. In D. J. Wess (Ed.). Paper presented at the Proceedngs of the 2009 GMAC Conference on Computerzed Adaptve Testng. Chen, P. (2011). Item replenshng n cogntve dagnostc computerzed adaptve testngbased on DINA model (Unpublshed doctoral dssertaton). Bejng Normal Unversty. [. (2011). DINA ()..] Chen, P., & Xn, T. (2011). Developng on-lne calbraton methods for cogntve dagnostc computerzed adaptve testng. Acta Psychologca Snca, 43(6), 710 724. [,. (2011).., 43(6), 710 724. ] Chen, P., Xn, T., Wang, C., & Chang, H. H. (2012). Onlne calbraton methods for the DINA model wth ndependent attrbutes n CD-CAT. Psychometrka, 77(2), 201 222. Cheng, Y. (2008). Computerzed adaptve testng new developments and applcatons (Unpublshed doctoral dssertaton). Unversty of Illnos at Urbana-Champagn. Cheng, Y. (2009). When cogntve dagnoss meets computerzed adaptve testng: CD-CAT. Psychometrka, 74(4), 619 632. Cheng, Y. (2010). Improvng cogntve dagnostc computerzed adaptve testng by balancng attrbute coverage: The modfed maxmum global dscrmnaton ndex method. Educatonal and Psychologcal Measurement, 70(6), 902 913 Cho, S. W., Grady, M. W., & Dodd, B. G. (2010). A new stoppng rule for computerzed adaptve testng. Educatonal and Psychologcal Measurement, 70(6), 1 17. de la Torre, J. (2009). DINA model and parameter estmaton: A ddactc. Journal of Educatonal and Behavoral Statstcs, 34(1), 115 130. Dodd, B. G. (1990). The effect of tem selecton procedure and stepsze on computerzed adaptve atttude measurement usng the ratng scale model. Appled Psychologcal Measurement, 14(4), 355 366. Dodd, B. G., Koch, W. R., & De Ayala, R. J. (1993). Computerzed adaptve testng usng the partal credt model: Effects of tem pool characterstcs and dfferent stoppng rules. Educatonal and Psychologcal Measurement, 53(1), 61 77. Dodd, B. G., De Ayala, R. J., & Koch, W. R. (1995). Computerzed adaptve testng wth polytomous tems. Appled Psychologcal Measurement, 19(1), 5 22. Hsu, C. L., Wang, W. C., & Chen, S. Y. (2013). Varablelength computerzed adaptve testng based on cogntve dagnoss models. Appled Psychologcal Measurement, 37(7), 563 582. Junker, B. W., & Sjtsma, K. (2001). Cogntve assessment models wth few assumptons, and connectons wth nonparametrc tem response theory. Appled Psychologcal Measurement, 25(3), 258 272. Kngsbury, G. G., & Houser, R. L. (1993). Assessng the utlty of tem response models: Computerzed adaptve testng. Educatonal Measurement: Issues and Practce, 12(1), 21 27. Leghton, J. P., Gerl, M. J., & Hunka, S. M. (2004). The attrbute herarchy method for cogntve assessment: A varaton on Tatsuoka's rule space approach. Journal of Educatonal Measurement, 41(3), 205 237. Mao, X. Z., & Xn, T. (2011). Improvement of tem selecton method n cogntve dagnostc computerzed adaptve testng. Journal of Bejng Normal Unversty (Natural Scence), 47(3), 326 330. [,. (2011). CAT. (), 47(3), 326 330. ] Mao, X. Z., & Xn, T. (2013). A comparson of tem selecton methods for controllng exposure rate n cogntve dagnostc computerzed adaptve testng. Acta Psychologca Snca, 45(6), 694 703. [,. (2013). CAT., 45(6), 694 703. ] Mao, X. Z., & Xn, T. (2013). The applcaton of the monte carlo approach to cogntve dagnostc computerzed adaptve testng wth content constrants. Appled Psychologcal Measurement, 37(6), 482 496. Rupp, A. A., Templn, J. L., & Henson, R. A. (2010). Dagnostc measurement: Theory, methods, and applcatons. Gulford Press. Tatsuoka, C. (2002). Data analytc methods for latent partally ordered classfcaton models. Journal of the Royal Statstcal Socety: Seres C (Appled Statstcs), 51(3), 337 350. Tatsuoka, C., & Ferguson, T. (2003). Sequental classfcaton on partally ordered sets. Journal of the Royal Statstcal Socety: Seres B (Statstcal Methodology), 65(1), 143 157. Tang, X. J., Dng, S. L., & Yu, Z. H. (2012). Applcaton of computerzed adaptve testng n cogntve dagnoss. Advances n Psychologcal Scence, 20(4), 616 626. [,,. (2012).., 20(4), 616 626. ] Templn, J. L., & Henson, R. A. (2006). Measurement of psychologcal dsorders usng cogntve dagnoss models. Psychologcal Methods, 11(3), 287 305. Wang, C. (2013). Mutual nformaton tem selecton method n cogntve dagnostc computerzed adaptve testng wth short test length. Educatonal and Psychologcal Measurement, 73(6), 1017 1035. Wang, C., Chang, H. H., & Douglas, J. (2012). Combnng CAT wth cogntve dagnoss: A weghted tem selecton approach. Behavor Research Methods, 44(1), 95 109.
140 47 Wang, C., Chang, H. H., & Huebner, A. (2011). Restrctve stochastc tem selecton methods n cogntve dagnostc computerzed adaptve testng. Journal of Educatonal Measurement, 48(3), 255 273. Wang, W. Y., Dng, S. L., & You, X. F. (2011). On-lne tem attrbute dentfcaton n cogntve dagnostc computerzed adaptve testng. Acta Psychologca Snca, 43(8), 964 976. [,,. (2011).., 43(8), 964 976. ] Wess, D. J., & Kngsbury, G. (1984). Applcaton of computerzed adaptve testng to educatonal problems. Journal of Educatonal Measurement, 21(4), 361 375. Xu, X. L., Chang, H. H., & Douglas, J. (2003). A smulaton study to compare CAT strateges for cogntve dagnoss. Paper presented at the annual meetng of the Amercan Educatonal Research Assocaton, Chcago. Zhang, Q. R. (2012). Cogntve dagnostc assessment preparaton and dagnostc studes on prmary school students chnese characters learnng (Unpublshed doctoral thess). Bejng Normal Unversty. [. (2012). ()..] Zhang, Q., & Ip, E. H. (2012). Generalzed lnear model for partally ordered data. Statstcs n Medcne, 31, 56 68. Exposure Control Methods and Termnaton Rules n Varable-Length Cogntve Dagnostc Computerzed Adaptve Testng GUO Le 1,2 ; ZHENG Chanjn 3 ; BIAN Yufang 2,4 ( 1 Faculty of Psychology, Southwest Unversty, Chongqng 400715, Chna) ( 2 Natonal Key Laboratory of Cogntve Neuroscence and Learnng, Bejng Normal Unversty, Bejng 100875, Chna) ( 3 Educatonal Psychology, Unversty of Illnos at Urbana-Champagn, Champagn, IL, 61820, USA) ( 4 Natonal Cooperatve Innovaton Center for Assessment and Improvement of Basc Educaton Qualty, Bejng Normal Unversty, Bejng 100875, Chna) Abstract Comparng to the nonadaptve testng, the major advantage of computerzed adaptve testng (CAT) s that the examnees acheve the same degree of measurement precson (.e., fxed precson). But few studes are devoted to the termnaton rules n varable-length cogntve dagnostc computerzed adaptve testng (CD-CAT). Inspred by the termnaton rule research n tradtonal CAT, ths paper proposed four termnaton rules for varable-length CD-CAT. The new termnaton rules were standard error of attrbute method (SEA), dfference of the adjacent posteror probablty method (DAPP), halvng algorthm (HA) and hybrd method (HM), respectvely. Then, the four new termnaton rules were compared wth the HSU and KL method under two scenaros: wth and wthout tem exposure control. Three exposure control methods were consdered,.e., smple, modfed restrctve progressve (MRP) and modfed restrctve threshold (MRT) method. The MRP and MRT methods were extenson of the Wang et al. s (2011) work to the varable-length CD-CAT scenaro. The results ndcated that: (1) When the crteron of varable-length termnaton rule was conservatve, the mean of the test length and the percentage of examnees reachng the maxmum test length were large, and the classfcaton accuracy rate for examnees who fnshed the CAT usng fxed precson was hgh. (2) Wthout the tem exposure control, the four new varable-length termnaton rules had a smlar performance compared to the HSU method. Wth the ncrease of maxmum posteror probablty and the decrease of, the classfcaton accuracy rate and the mean test length presented a ncreasng trend. But the tem pool usage was unsatsfactory. (3) Wth the tem exposure control, tem pool usage was greatly mproved n the sx varable-length termnaton rules whle the classfcaton accuracy rates were mantaned. Dfferent exposure control methods had a dfferent effect on the dfferent varable-length termnaton rules. The relatve crteron termnaton rules such as DAPP and KL methods were easly affected by the tem exposure control. (4) Taken all together, the SEA, HM, and HA methods were comparable to the HSU method, and followed by the KL and DAPP method. Some future drectons were suggested n the end of ths paper. Key words cogntve dagnostc computerzed adaptve testng; varable-length termnaton rule; exposure control; classfcaton accuracy rate; DINA model