!
!
!
!
"
"
"
"
#
#
#
#
$
$
$
$
! " # $ % & ' " $ ( % ) & * + , - . / 0 1 2 3 4 , 5 67 8 9 : ; <=> ? @ , A . / B C D;=4 E , F G H I J K $ 9 L M N O P - Q R S PT U V W X Y Z [\] 3 ^ X _ ` a b c d e f g h ij k l m n n o p N O qr s t u v w x y z { | J } x ~ q , - 5 d e ~x % < <R S \N } wx [~ * + 5J d e g ~ Y Z wx 9 = ¡ ¢ £¤ 5¥ ¦ d e T U V § q - Q £W X * + ¨ 6© ª £« ¬ wx ~@ , 7 8 q§ ® < ql ¯ °± ;² 5³ ´d e µ ¶· ¸ ¹ º » 7 £N O wF G 7 8 4 ¼ @ ½ 5 * ¾¿ À Á  J ¾ÃÄ Å Æ Ç =È É v Ê Ë =Ì Í Î Ï Ê 7 8 @ , _ Ð Ñ Ò Ó Ô © Õ Ö × £Ø Ù Ú Û Ü Ý Þ ß W X à ~T U V £á 7 â ã ä W ~å æ ç è <é ê ë C ~ì í î ï ð ñ [\¨ 6~msn + ò ó ô õ £^ X ö ÷ ~ 7 8 @ , F Ý ø ù ú ú û ü 5F * qý þ Pq 7 8 @ , J Ø A § qÞ ß © Ø =Ö ³ ÷ 5 º [67 8 x ³c _ 2011.06
á standard LRT method, LRT-STP¼ ² u LRT method with scale purification, LRT-SP[\ pure anchor ! " "# LRT method with pure anchor, LRT-PA$% & © ' ( ) * + likelihood ratio test, LRT#~$, - . °[. / ¤ 0 1 ¶"2 4 differential item functioning, DIF+ 3 4 4 * 5 5. / 67 8 9 % : B ; < = é >? ¶@ 4 ;é A ability differencePB U DIF¶"C é * DIF percentage[\DIFD E DIF patternFG < = D H0 Type I error£+ ;power5
Ø Ù I J ~ë & . / ð K °% & ¤ #DD H0 L 1 M N ~O P
Q R S 5+ ;¤ ã % & ¤ #~DIF + 3 + 4 4 T é U æ V L q DIF ¶"C é * W ¼ =°X Y Z 5 é [ ~DIF D E constantV \] 4 ;ó $ð ^ °LRT-SP#D_ü ` a 9 LRT-PA#= DIF¶"C é * Ô W ¤ #D+ ;L 1 ` b9 LRT-ST#Fc\] 4 ;© ó $V DIF ¶ "C é * W ¼ d 30%<LRT-SP#e f ü Íg b+ ;5¤ ã cDIF D E balanced<© 7 \] 4 ;ó $£h Ø Ù i ü % & ¤ #L 1 j ü k l + 3 4 4 5 m n Ø Ù N Í~GRM. °o ( ) * + # DIF+ 3 R S ~+ 3 ¼ ² u p DFTDq ` r s ~bDIFC é * B U 1 t u v+ 3 4 4 W ¼ + 3 Ø Ù w x 5 y z Ñ {0 1 ¶"2 4 P( ) * + P² u P$, - .
A Comparison of Three Likelihood Ratio Test Method in
Assessing Differential Item Functioning for Polytomous Items
Abstract
Three different DIF assessment methods, “LRT -ST”, “LRT -SP” and “LRT -PA” was compared in assessing DIF under the graded response model. Three independent variables were manipulated in the simulation study , including the ability difference of subjects, the percentage of DIF items in the test and the DIF patterns. The dependent variables were Type I error and power of DIF assessment.
The results showed that in a variety of simulated situations, Type I error rates of three methods were well-controlled. The difference in power between three methods was pretty small, and power rates decreased while the DIF percentage increased. Furthermore, when the DIF pattern was constant and the mean abilities of the two groups were equal, the LRT-SP performed better than LRT-PA. With the percentage of DIF items in the test increased, the other two methods yielded slightly higher power than LRT-ST. When the abilities of the two groups were unequal and the DIF
percentage increased up to 30%, LRT-SP method showed higher power than other two methods. However, when the DIF pattern was balanced, all three methods performed similarly under every conditions in this study. It was recommended that the scale purification or DFTD strategy should be included into the LRT method when assessing DIF for polytomous items.
Keywords: differential item functioning, likelihood ratio test, scale purification, graded response model
_1 ¶" ... 19 _2 \] 4 ;ó $V DIFD E constant ... 21 _3 \] 4 ;© ó $V DIFD E constant... 22 _4 \] 4 ;ó $V DIFD E balanced... 23 _5 \] 4 ;© ó $V DIFD E balanced ... 23 _6 \] 4 ;ó $V DIFD E constant1~4°D+ ; ... 24 _7 \] 4 ;© ó $V DIFD E constant1~4°D+ ; ... 24 _8 \] 4 ;ó $V DIFD E balanced1~4°D+ ; ... 25 _9 \] 4 ;© ó $V DIFD E balanced1~4°D+ ; ... 25 _10 D H0 < 1 é [ Ø Ù ... 27 _11 + ;< 1 é [ Ø Ù ... 28
~ü , B U é ê » ² Ø Ù P4 ;f ü =, $ G C ¡ ¢ N £ % ¤ ¥ B Pä B ¦ © r 65§ 9 B U Ø Ù $ D¨ © ª 6B U S « Ö h 4 ¬ B Í? ¶@ 4 ;B U Ø Ù Ö h ® qO P ¯ 4 £ [\° x ± ± , ä ² y H S "5=[ ³ B U p² _´ µ <¶ Ç · B U ¯ £ P4 £ D é [ ó 5 D°9 B U é ° x D 5 ¸ 5¹ =æ ó y é [ ¡ ¢ } x 9 B U ° x ó y S "D º q» W ¼ DY Z 5 ¼ ½ B U ° x ¾ ¿ À Á DÂ Ö B U » ² § q? ¶@ =À Ã Ä ® q| x w§ q? ¶@ ù 4 Å q $» Æ Ç 5V F È » Æ © ? % $ ñ < = p§ É \] ¨ © r Ê \Px >P] Ë Ì Í Î $ r 6¤ 4 w Ï Ð B U é DÑ 7 Ò w £h Ó Ô ¯ B U Ø Ù ° x 5= ü B U L ã å f B 9 Õ \] ð ^ 6B U 9 © ' ñ < = \] =À Ö h ® qó 'B ² Á Ö pB ² Ö × 3 Ã Ä y H á · 7 S "5 " + Ø Ù Ú Û Ü 9 4 ;ó 'A Ý É © '\] ? ¶@ ~'Þ ¶ "+ q© ° é ð ß e à Ð Þ ¶"® q0 1 ¶"2 4 differential item functioning, DIF5á ¶"- k 7 item response theory, IRT°ó y é [ ¤ #× 3 â 1 á ¤ #Y Z + 3 ² _pB U ¶"Ö h ® qDIF5 =B U ´ µ @ =À G B U ã é ¶"äÃ#q4 - Í? ¶@ ~ B ²
å + a æ £ e B U ç ] £ ¡ ç ] â è é F È ¶"u Íê ë Ó
ì pí î 5
± 0 1 ¶"2 4 + 3 ó y S "67 ê [ é . ï ð ñ ñ ò Ç =ü § ó ô » ² põ ö E £ P$² _w? ¶@ » ² Ø Ù ÷ Ã# ¿ ø é pù é ¦ _J =? ¶@ ú û ü #pÖ ý 'þ - £ © 1 ¿ u , qpÃë 5 ú T + "~ u [\ B ² ( ¤ Y ~t x ä D B
U r TIMSSThe Trends in International Mathematics and Science StudyP
NAEPThe National Assessment of Educational Progress$æ ¦ ÷ ö r "P¿ "$R - ¶"56r ¼ ~ é "p / é
"D B U + 3 0 1 ¶"2 4 I · 7 5
~ qy é "DIF+ 3 ¤ # 1 é / IRT [\IRT 67 á IRT ã é 5õ á DIRT DIF+ 3 ¤ #q( ) * + # likelihood ratio test, LRT; Thissen, Steinbreg, & Wainer, 1988\¶"£0 1 2 4 + B #differential functioning of items and tests, DFIT; Raju, van der Linden, &
Fleer, 19955 ( ) * + #1 ~ é "£ é "DIF+ 3 5á 8 D 1 i ü ± ~ é "DIF+ 3 ( ) * + # ð ß ó c Bolt, 2002; Cohen, Kim, & Wollack, 1996; Kim & Cohen, 1998; Stark, Chernyshenko, Oleksandr & Drasgow, 2006; Wang, 2004; Wang & Yeh 200367 LRT#1 [ + 3 © '& DIFÂ | x DIF\/ | x DIF[ ° 5 ~ é " ¤ ã @ º i ü ~$, - .
° ( ) * + #+ 3 0 1 ¶"2 4 <~D H0 M N + ó ck l
Ankenmann, Witt, & Dunbar, 1999; Kim & Cohen, 1998; Bolt, 20026 ö Thissen$3 ~1988§ u Í( ) * + #¦ é "DIF+ 3 5
§ 9 ~ DIFé [ <äB U qDIF¶" 1 4 O | $ ! HV DIF¶"* "bP# \] q$ "b! Hð ß "¹ · = ¨ © DIF+ 3 Ø Ù Lord, 198056 cB U DIF¶"* % 10% < DIF+ 3 ¤ #DD H0 Type I errorù i $ & ' inflationü (~D H0 & ' ù M <DIF+ ;powerâ © 1 ¯ 5 ) 6* " q @ u Ͳ u scale purification ( Candell & Drasgow, 1988;
Clauser, Mazor, & Hambleton, 1993; French & Maller, 2007; Hidalgo-Montesinos & Gómez-Benito, 2003; Holland & Thayer, 1988; Lord, 1980; Park & Lautenschlager, 1990; Miller & Oshima, 1992; Navas-Ara & Gómez-Benito, 2002; Wang & Su,
2004V á . / i ü ~DIF¶"© % 20%ð ß °² u ~© ' ¤ #+ 1 [E % q4 M N D H0 4 Ù 5© cB U DIF¶"* % 20%<Â ² u D H0 G + i $ & ' =ù M ð ß 5 6 "#constant-item method, CI ;Thissen, Steinberg, & Wainer, 1988; Wang & Yeh, 2003( u ÍÓ V i ü ~ 6¤ # DIF+ 3 <ä4 w , qDIFDIF-free¶"- ! "anchor item 1 q4 M N DIF+ 3 D H0 W ¼ + 3 4 4 6Â Wang2008§ u Ía ! ´+ 3 DIF-free-then-DIF¿ à DFTDq ` 5á . / WangP SuniShih2010 i ü DFTDq ` î q4 M N DIF+ 3 D H0 Ó 1 ` . u v+ ;5
m n + Ø é [ / IRT ( ) * + # á ~ é . ° ² u q ` Ö h 4 r [³ ~ é § i ü 1 [q4 M N D H0 / cB U DIF¶"* ® qbC é * <² u 4 4 Ö h º 4 q4 i 0 5 î ¼ ² u ( ) * + #~
º ¼ ( ) * + #[\ DIF-free¶"- ! " "#[6% & ¤ # . / Ó * 5 6% & ¤ #~ é . °
cG B U 1 : \] ? ¶@ f B V @ * 5 © '\] r x >P& Ê P Ì Í Î P2 3 4 $? ¶@ ~6G B U + _ü Ö h q§ 0 1 < @ Ã Ä a w Ï Ð B U § B ² % Á 5 p( \] =À Ö h ó '5 c? ¶@ ¥ 9 \] p ñ 6 © '~¶"+ - + $ 5 qp5 © ð ^ e Ð ¶"â ® q§ ½ 0 1 ¶"2 4 5 é "=À Ö ¶"s 7 é £? ¶@ 8 ~4 ;D9 y : c4 ;ó 'A Ý É 9 © ';] \] ~¶"s 7 é + < © ó $<J Ã#6\] ? ¶@ B U Íó '8 ~< = =6¶"Â = ® qDIF ü (Cohen, Kim, & Baker, 1993; Kim & Cohen, 19985
0 1 ¶"2 4 G x 1 é | x uniform\/ | x nonuniform& 5ä[\] - < ¦ > * 5 c ? \] reference group~# ¶"+ - | 2 9 qÍ Î J Ö © 7 Ö b4 ;P$4 ; @ Ö A 4 ; ? \] ? ¶@ * B Þ ;] focal group? ¶@ ® q5 bp 5 A < e à Ð ¶"® q| x 0 1 ¶"2 4 uniform DIFF ¤ ã ä9 b4 ; ? \] ? ¶@ =À ~# ¶"+ - * B Þ \] ? ¶@ ¦ qA 9 A 4 ; ? \] ? ¶@ =À e * B Þ \] ¦ © 6<© '4 ;4 9 ? ¶@ \] q$ © | <â à Ð ¶"® q/ | x 0 1 ¶"2 4 nonuniform DIF5 ¤ â C @ D E & D DDIFÁ 5 × 3 1 " H k ² F Þ á ¶" item characteristic curve, ICC¦ Ù Ú © '4 ;P© '\] D
? ¶@ ~'Þ ¶"+ < 5ä ? \] £B Þ \] D- - é > Ó [G µ \] DICC 9 ' ã + e ® q| x 0 1 ¶"2 4 ¶"1 _J r 15 1 | x 0 1 ¶"2 4 § 11 ü ͧ 9 'È 7 ~6¶"E % 0.5 < B Þ \] § è 4 ; 5 bI J 6¶"9 B Þ \] ? ¶@ Ö 5 H I º  9 \] ? ¶@ =À 6¶"bI £ Ú I q§ © '5 F J £XK L M £\] ICC ë C ó N § < é >~0.7£0.5O P Q _ Á 5 Ö 9 ó '4 ;6R 4 ; S 0.55? ¶@ =À ? ;] * B Þ ; ] ~Ð "+ 0.2 < 6Â Ø D2 9 qÍ Î 5V © 7 ? ¶@ 4 ;bA q\] L ? \] 6<e à 6¶"® quniform DIF5
¤ ã r Ù 9 b4 ;£A 4 ;? ¶@ =À 2 9 qÍ Î \] © 'º Â r °2§ J \] ICC $ N T Dð ß J Ö aUV >£ W q0 1 6<q\] $ £? ¶@ 4 ; qy =/ r 'Ø | uniform DIF 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 .0 - 3 .0 - 2 .0 - 1 .0 0.0 1 .0 2 .0 3 .0
x q9 # \] <Â à Ð ¶"® qnonuniform DIF5 2 / | x 0 1 ¶"2 4
é . ~c? ¶@ - ï ð ÷ Ã#X ¿ ø Yp 'Á © 'Á ¦ j ü =Ö q & 1 4 x - B U p² _ r ¿ "PZ 7 "$[ \ - "D [\] ^ _ ² _Likert-type scaleP3 ` ² _ $E £ ² _5 ~ Ö á é > © 'é p £ © '$, ¦ Ï q? ¶@ - - @ a x 5æ 2 T ¦ b ÷ i f Í & é . ¦ é [ © '
é ï ð õ c é . q$, - . Graded Response Model,
GRM; Samejima,1969P» ² . Rating Scale Model, RSM; Andrich, 1978Pã G é . Partial Credit Model, PCM; Masters, 1982[\ u ã é é . Generalized Partial Credit Model, GPCM; Muraki, 19935 & . ¿ ø r °{ P $, - . nonuniform DIF
§Samejima1969§u ÍdC é . 5 9 é [ q ? x - - ï ð r] ^ ² _ pÖ e C P k ; » ² Reise
& Yu, 19905 ~ ? ¶@ - ñ ò < è 7 f º = Ö 9 U indirectIRT. Embretson & Reise, 20005
a Í4 ; θ? ¶@ ~ > g βij° 8 - Pix∗(θ)5 i ij i ij i ix a x j m a P 1,..., )] ( exp[ 1 )] ( exp[ ) ( * = = − + − = β θ β θ θ 1
6 8 - operating characteristic curves, OCC ∗(θ)
ix
P Â h i <
º _ J 4 ; θ? ¶@ ~ i" % x é [ + < 5 ai i" V >
£ βij i" j : = g mi g: mi+1 - > :
r{q j : - = Â q : I £ 5
U é > k > - l category response functionº Â 4 ; θ ? ¶@ ~ i" % x é < 5 ( ) ( ) * ( ) 0,...,4 ) 1 ( * − = =P P + x Pix θ ix θ ix θ 2 m Ö *( ) 1 0 θ = i P V *( ) 0 5 θ = i P 6 1 k Í~ j : é ¶" ± : > - < r° { 0 ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( 1 ) ( ) ( ) ( * 4 * 5 * 4 4 * 4 * 3 3 * 3 * 2 2 * 2 * 1 1 * 1 * 1 * 0 0 − = − = − = − = − = − = − = θ θ θ θ θ θ θ θ θ θ θ θ θ θ θ θ θ i i i i i i i i i i i i i i i i i P P P P P P P P P P P P P P P P P 3 P » ² .
² _  ? ¶@ ~ ë» é Þ + 0 1 §q " ~ = À ù Ö ó ' 5 o p 6 q ë¶" r Õó ' g Vs N g D i 05 [ 4 ; θ? ¶@ = À ~ i" + x é < 1 §° t ° 4 _ J {
∑
∑
= = = = − + = ∑ − + ∑ − + = 0 0 0 0 0 [ ( )] 0 ]} ) ( [ {exp } )] ( [ exp{ ) ( j i j m x x j i j x j i j x P θ λ δ δ λ θ δ λ θ θ 4 λ i i" I £ δje Ö ² _ i" j : g 5 %P ã G é . ã G é . Ö §Masters1982u Í a ~ Ö u o ø pø YF é # © ° ü ( w ã é ¶" ? ¶@ 1 q ã é é 5 A i f d ~ v Õ 7 Ö " ~ » é Þ q ? ( % A é 5 « % b é 5 H I < 4 9 6 & . 5 4 ; θ? ¶@ ~ i" + x é < 1 §° t ° 5_ J {∑
∑
= = = = − ≡ ∑ − ∑ − = 0 0 0 0 0 ( ) 0 ] ) ( [exp )] ( exp[ ) ( j ij m r r j ij x j ij ix i P θ δ δ θ δ θ θ 5 δij i" j : f I £ 5 P u ã é é . 6 . Muraki1992 dPCM. = ¦ 5 w ' G ² ë¶" q © ' V > £ x ~ 5GPCMÖ & ® q y x . c B U ë¶ " a ó ' < , PCM. Fä ¥ s N ë¶" r ' ' Õg < RSM. 5GPCM. 1 §° t ° 6_ J {( ) 0 )] ( [exp )] ( exp[ ) ( 0 0 0 0 0 − = − ∑ − ∑ =
∑
∑
= = = = j i ij m k i ij k j ij i x j ix a b b a b a P θ θ θ θ 6 aiÖ N i" V > £ bij i" j : I £ 5" 90 Q z » ² ¤ » > u < B U " D © ¥ Õ Ö ø ) é " ß [\pR-»²{®[| 6 5 »²{®=À ¯ 4 £P° x $ N G Ç · 7 5ú T +ü _ü»²¤ *z ± u "DBU J « Í üq! H * "56~ é PR-¶ "» ²+BDIF I · 7 5 [³ ~ é . î i f Í ó c DIF+3 ¤ # J h i , ÙSwaminathan & Rogers, 1990; Zwick, Donoghue, & Grima, 1993;
Clauser,Mazor, & Hambleton, 1993; Rogers & Swaminathan, 1993; Fidalgo,
Mellenbergh, &Muñiz, 20005æ ¦2 & º i f V ~ é "DIF +3 +5~ é "DIF+3 ¤ #ä | +1é & {/IRT [\IRT 5 @M U [F J é -A<= matching variable é ó ' ?¶ @ 4;ó ' F´ @eG ?¶ @4; -A<= +3 4;ó ' ?
¶ @~' Þ ¶ "+ < p¶ " Ö h ó $ 6Ã Ä a ¶ "
\4; 5¶ Ç IRT ¤ #} \¶ " £4; ~ +5/IRT ¤ #¦ ~ <A Ò £5b6ê @ § ö 5
Mantel-Haenszel#GMH; Mantel & Haenszel, 1959Ppoly-SIBTEST#Shealy & Stout, 1993P 4 l é [#logistic discriminant function analysis,LDFA;
Miller & Spray, 1993\ordinal logistic regressionOLR; Zumbo, 1999F=IRT
+3 ¤ #q( ) *+ #likelihood ratio test, LRT; Thissen, Steinbreg, &
Wainer, 1988\¶ "£012 4+B#differential functioning of items and tests, DFIT; Raju, van der Linden, & Fleer, 19955" 8 ~$ , -. ° ( ) *+ #+3 01¶ "2 4<~DH 0M N+ó ck l Ankenmann, Witt, & Dunbar, 1999; Kim & Cohen, 1998; Bolt, 20026 ö ( ) *+ #¦ é "DIF+3 5
( ) *+ #³ Ö § Thissen$ 3Thissen, Steinberg, & Gerrand, 1986;
Thissen, Steinberg, & Wainer, 1988, 1993 9 DIF+3 +5[% IRT. LRT# ¶ "DIF+B<6 7 [°%: f {
P \ ] =À s NBU § q¶ " Ló ' º Âq § q¶
" , qDIF6§ ½ . compact model5D´ G °t ° 7 k . likelihood devianceÓ[GC2_J D{
GC2 =−2×log(likelihoodcompact) 7
P é +3 D¶ "studied item î ¶ " \ ] ó $ Ds N º « Ð ¶ " \ ] ?¶ @ ©ó ' õ N a pb 6 § ½ d . augmented modelU ' È k likelihood devianceÓ [G _J D{A2
G2A =−2×log(likelihoodaugmented) 8
%P k d . £ . Dlikelihood deviance01Ó[G _J D2
º ÂG2 G2 G2
A
: D05rr~ f X « ¶ "b ó $ eC § £1Fä' < « ¶ "a £b \ ] ©' eC § £25äG ä 9 DC § £2
¤ eLRT# +3 ¶ "®qDIF5
[G ®q30Þ ¶ "V£ . 5 ABU 6<ä7 +B1 "Ö h ®quniform DIFe` aR; . s NBU § q¶ " \ ] =À Ló ' º Âa G BU ÓÃDIF¶ "Ó k G FU ~d . C2 î 1"I £ \ ] Ló ' Ds NAêq V > £ \ ] = À Ö ó ' 6< k G F³ ´ k2A 2 C G £ 2 A G D0ÂG § 9 +Ø . 2 X ó 0: ¶ " 1"I £ G2 " C § £1 ¤ é A =) ª 2 3.84 ) 1 ( = X ä k ´ DG ä 92 3.84e 1"®quniform
DIF5r +3 ¶ "~V > £ +Ö h q§ 01º ÂÖ h ®qnonuniform DIF <Õ è 7 ~ f q ¶ " \ ] =À V > £ ©' AI £ ó ' eG ' È " C § £2 1 ¤ é A) ª º 2 3.84 ) 1 ( = X Fä @
Õ +3 ¶ "Ö h ®qDIFA©~Á Ö ¼ & DIF<eX è ~ f q 1 "I £ \V > £ \ ] L©' 6<G " C § £2 2 ¤ é A) ª 2 5.99 ) 2 ( = X äG ä 92 5.99e Ð ¶ "®qDIFd9 Ð "Ö
uniform DIF pÖ nonuniform DIFe é > é a \b +3 5
DIF
~ DIF+3 © IRT p/ IRT +3 ¤ #LÄ á R; r ' ² common metric [©' \ ] ¶ "-> *55 r ' ² ÂÖ § ½ A<= matching variable~ DIF+3 <Ã Ä w
Ï A<= Ö P©®DIFü( ¶ "äBU q DIF¶ "e ² ? contamination =O | 4; ! H [\© w +BØ Ù5=[³ LN Í cBU ®qbC é *DIF¶ "<DH 0Â « i $ & ' ù M ü( 6<+BÍ ØÙâ ©1¯ 5 ) 6* "q
@u Í ² u scale purification ( Candell & Drasgow, 1988; Clauser, Mazor, & Hambleton, 1993; French & Maller, 2007; Hidalgo-Montesinos & Gómez-Benito, 2003; Holland & Thayer, 1988; Lord, 1980; Park & Lautenschlager, 1990; Miller & Oshima, 1992; Navas-Ara & Gómez-Benito, 2002; Wang & Su,
20045~ Ö BU ®qDIF ü( ¶ "C A<= î ¥ [F t à DIF ¶ "Õ , DA<= BU ¶ " DIF+3 5
± 8 LI J ~DIF+3 ¼ ² u 1q4 X A DH 0Óu v+3 4 4rFidalogPMellenberghiMuniz2000~3PLM° i ü~DIF¶ "C é *10%[°<MH#1qP _üAcDIF¶ "C é *W ¼ 15%< ² u MH-2£ Q ² u MH-iê4q©Y _ ü=ä¥ DIF¶ "C é *u vd30%<Õ °MH-iê4 $ P ØÙ5= ~SuiWang2005 ' È J i üMH-2PMH-i4 Ùa 9 MH5 ¤ ã ~ é "DIF+3 +Wang iSu2004b N Í ~Mantel#\GMH# ¼ u J q' È 4 Ù ² u Mantel-2£ Q ² u
Mantel-i4 Ùa 9 Mantel# ² u GMH-2£ Q ² u GMH-i 4 Ùa 9 GMH#5 J q N Í LR#+B é "<~ä È \DIF ¶ " ²bð K °¼ u ' È 1[ q4 Í +BÍ uniform DIF
Hidalgo-Montesions & Gómez-Benito, 2003; Su & Wang, 20055
Ç =T +² u X ~DIF ¶ "C é *5A <4q5P 4 Ùc
G BU q DIF¶ "<Ââ ¼ u DH 0ê i $ & ' ù M ü( 5u o 6ü( "#constant-item method ,CI; Thissen, Steinberg,
& Wainer, 1988; Wang & Yeh, 2003( u Í ± º ÷ T 6#*ó $ I £#equal-mean-difficulty, EMD\ "#all-other-item, AOI+ B4 Ù Wang, 2004; Wang & Shih, 2010; Wang & Yeh, 20035Vc " # DIF+3 <ä4a w Õ , q DIFDIF-free¶ "- ! "anchor item¥ é BU ¶ " +3 1q4 M NDH 0 W¼ +3 4 4Candell & Drasgow, 1988; Cheung & Rensvold, 1999; Stark et al.,
2006; Thissen et al., 1988; Wang & Yeh, 200356ÂWang2008§ u Í a ! ´ +3 DIF-free-then-DIF¿ à DFTDq ` DFTD-#â Ö á Í ³ ©14®qDIF ¶ "-DIF-free¶ "[6 ! " "#DIF+B ! " RS ö Þ ¶ "Wang & Yeh, 2003; Shih &
Wang, 20095á . / Wang$ 32010i üDFTDq ` î q4 M NDIF+3 DH 0Ó1` . u v+ ;5
6 7 ~ á ~ GRM. ° IRT ( ) *+ #á ¼ ² u q ` Ö h 4~®qDIF ¶ "BU q4 M NDH 05 î ¼ ² u ( ) *+ #LRT method with scale purification, [ °¿ à LRT-SP# º ¼ ( ) *+ #standard
LRT method, [°¿ à LRT-ST#[\ pure anchor DIF-free¶ " ~. /ï ð ÷ ÃDIFü( D¶ "- ! " "#LRT method
with pure anchor, [°¿ à LRT-PA#§ 9 LRT-PA#Ö [w = ÃDIF ü ( D¶ "- ! "r s DIF+3 DDH 0M N q³ k l 4 Ù 6 ØÙ1-*5 á ¤ #+3 4 4D 5 [+%& ¤ # . / Ó*5 á 6%& ¤ #~GRM . °DDIF+3 4 45 [°¿ ø LRT-SP#[\LRT-PA#D+3 f 5 LRT-SP#+3 f r°{ 1BU § q¶ " DIF+3 F 2 +3 Í DIFD¶ "C A<= î F 3· BU ë ¶ " DIF+3 F 4· f 2£ f 3M % ó ?DIF+3 ØÙó ' v 5 LRT-PA#6 7 Ö ~. /ï ð [" Ã DIFü( ¶ "- ! "G ( ) *+ # "# ¶ " DIF+3 5
t 9 _ 1§ -@ Matlab $ GRM. D. /ï ð ´ ¥ [IRTLRDIF DIF +3 5
6 7 89 %: B ; <= é > ?¶ @4;é Aability
differencePBU DIF¶ "C é *DIF percentage[\DIFDE DIF pattern5 P ?¶ @4;é A 6ã é 6 7 89 & ð ß é > \ ] 4;ó $ £ \ ] 4;©ó $ 5~ \ ] 4;ó $ ð ß ° ? \ ] £B Þ \ ] 4;L89 0< 11 õ E é AF~ó y L B Þ \ ] Z \ ] 6~ \ ] 4;©ó $ ð ß °e ? ; ] 4;£+Ø ó ' =B Þ \ ] 4;89 -1<11 õ E é A5 P BU DIF¶ "C é * ~ r 89 j & ©' £DIF¶ "C é *é > 0%P10%P20%P 30%[\40%_J 89 DIF "G 0P2P4P6[\8"m [³ 1i üDIF ¶ ""W¼ « , DH 0& ' ù M ü( Finch,
2005; Wang & Yeh, 200356 /á 89 bE40%DIFC é *[é > LRT-ST#PLRT-SP#\LRT-PA#~bC é *DIF¶ "BU DH
0M N4 45
%P DIFDE
6 7 89 & ©' DIFDE {@ ø $ constant\ ¡ $
balanced5constantDE § qDIF¶ " ? \ ] q =balanced DE eÖ ¢ DIF¶ " ? \ ] q ¢ e B Þ \ ] q 5 £ ~¬ T ð K °DIF+3 4 4_üQ R 89 ³ ¹ · \³ . & DIFDE 5§ 9 j Þ ² q: £¤ n ¬ T ð K º ~© ' £+§ , DIFð ß Ö h ¨©+3 4 4 ~ & DE °é > 89
"Â qI £+01 "e qI £+01=ä89
DIF"W¼ "%"e %qI £+01"e ùqI £+01[6-¥ Ñ 5 ~balancedDIFDE °ä89 DIF ""<e ; ] ë 89 "Vé > £ qI £+015
î +Ø 89 <= ~ ¤ ã § È ? \ ]
10003B Þ \ ] 5003BU - £20"5=~ BU uniform DIF5 õ c DIF D6~ X á uniform DIFð ß ~DIF¶ "I £8 9 + \ ] ~DIF¶ "I £01L " 0.3\ 00.05õ E é A$ £DIFü( V§ qð K °ï ð L· . /100?5
G <= DH 0Type I error£+ ;powerDH 0Ö ÃDIF¶ "Y H N = , DIF¶ "ð ^ =+ ;e w N = Í
DIF¶ "ð ^ 5 § I ¦ .05Ì § = é A k ~100 ?· T U 1« 0d.0927 < ÃDIF¶ "H DIF¶ "6n k 1U ?4 9 Q R äDH 0 9 & ' % Í Q R D 6<Ââ q¥ b+
;J ©®qÁ Ö 5
~ ± DIF¶ "C é * £õ § ¨ DH 0ù M ¨ ©+3 4 46 A~WangiSu2004a e= average signed area
[°¿ à ASAä © ª Ö , DIF+ #ù M Ã 7 5ASA \ ] ~ s 7 é é A+expected score distribution§ $ ã i 5 ° r°{
SAi =(1−ci)(biF −biR) 9
SAii"signed areacii"« B£ biFibiReé > B Þ \ ] i ? \ ] ~i"+I £ 5ASAÂSA ° {
ASA SA I ci biF biR I I i i I i∑1 / =∑1(1− )( − )/ = = = 10
I BU "5cci0<6< . °ASA° r°{ iF iR F R I i b b I b b ASA=∑ − = − =1( )/ 11 " ° 111~ . °ASA \ ] I £D056 cASA0<Q _ \ ] I £ó ' DIF¶ " \ ] =À Ó¬ ß , q p@© ð ß º _J ¶ qDIF¶ "Aa ] =À \ ] ¨©Ö ° 5 -DcASAä 9 0<Q _B Þ \ ] I £ä 9 ? \ ] I £ Â_J DIF¶ " ? \ ] ß , 5q ð ^ 5~Wang2001 N Í ~
balancedDIFDE § k Í ASAT é U æ 9 0F=~constantDIF DE cDIF¶ "C é *20%ASA0.09< "# DIF+3 DDH 0@ 1M NO P Ç =ä~ó ' ð K °ASAu bd0.18< e O | DH 0 $ & ' ù M ð ß F ¤ ã cASA 0<Ââ BU DIF¶ "C é *u bd50%DH 0ê1Ï ¯ ~M NQ R DS5=Wang iSu2004a º = cASA©0<Mantel#iGMH#~+BDIF
¶ "+DH 0 qù M ü( 5§ 61c î DIF¶ "C é * £
ASA Ö ¨©DIF+3 · 7 N 56 /~± & ð K °é > k
_1 ¶ " Item a b1 b2 b3 b4 1 1.57 -0.38 0.49 1.04 1.67 2 1.31 -0.61 0.63 1.37 1.82 3 1.63 0.01 0.67 1.33 1.91 4 0.97 -0.23 0.31 0.98 1.62 5 1.53 -0.31 0.60 1.27 1.79 6 0.99 0.06 0.99 1.50 2.02 7 1.79 -0.10 0.35 1.01 1.72 8 1.39 -0.02 0.63 1.28 1.87 9 0.85 -0.35 0.67 0.97 1.24 10 1.20 0.18 0.70 1.29 1.42 11 1.05 -0.37 0.03 0.54 1.01 12 1.42 -0.56 -0.13 0.67 1.54 13 1.23 -0.36 0.53 1.20 1.52 14 1.34 -0.52 0.39 1.54 1.89 15 1.63 -0.53 -0.12 1.27 1.81 16 1.45 -0.20 0.49 1.00 1.59 17 1.71 0.02 0.56 0.93 1.48 18 1.28 -0.44 0.18 0.72 1.31 19 1.43 -0.01 0.39 1.36 1.69 20 1.10 0.10 1.06 1.61 2.01
IRTLRDIFÖ § ® ¦¯ ä University of North Carolina at Chapel HillDavid Thissen m § 6 ] [( ) *+ #k 7 ¥ ° á d . 89 IRTLRDIF1 [ uniform\nonuniformDIFé [± [
IRTLRDIFÖ ² ³ ~ ] 1w qÁ ´ DIF+3 C @á µ * °Û ¼ + ¶ · ó c 69 C @~ ] \ + ó c¤ â 5
IRTLRDIF DIF +3 <1 "#p "#5 ] r ¸ Ö ö "#º Âî +3 ¶ " BU ¶ "Lq Ó¬ ®qDIF5 Ç =~WangiYeh2003 i ücBU ®q5bC é *DIF ¶ " <¨©6#D+ ;56º q @ ¹ "#î ~DH 0 M N+1[y 5 4 Ù ! "W¼ + ;J Du v5 Ç = ²üT ð ^ ö ! "W¼ ! " DIF¶ " < 6 =À ¸ RS ö 4Þ ¶ "- ! "Thissen et al., 1988; Shih & Wang,
2009; Wang & Yeh, 20035
IRTLRDIFÕ 4 ¦+3 ¤ n é p% . [\ é GRM. ï ð Ã# 9 Rasch. p é . 6~ +è 7 g¼ º Á 5
DØÙé > t 9 _ 2d_5_ j ü%& ¤ #~©' ð K ° DH 0£+ ;[°ØÙé DIFDE Ù Ú5 PDIFDE constant _2 \ ] 4;ó $ ØÙ~ë DIF¶ "C é *ð ^ °%& ¤ # DDH 0L1?% O P M NLRT-ST #~bDIFC é *<g. +vd.04 =LRT-SP#£LRT-PA#eÔ ¯ ~.02 d.03 D9 5 + ;¦Ù Â~bDIF C é *ð ^ °%& ¤ #_üL~.84[+=DIF¶ "C é *W¼ % & ¤ #D+ ;ùqggÔ Y Z 5a ] +=À LRT-SP#£LRT-PA#_ üT é U æ V~b DIF¶ "C é *ð ^ °ù5 LRT-ST#¦ P t 5
_2 \ ] 4;ó $ VDIFDE constant
DIF% ASA
Type I error Power
L-ST L-SP L-PA L-ST L-SP L-PA 0% 0.00 .03 .02 .02 10% 0.04 .02 .02 .02 .99 1.00 .97 20% 0.15 .03 .02 .02 .85 .86 .84 30% 0.19 .03 .03 .03 .87 .91 .90 40% 0.30 .04 .02 .03 .84 .90 .90 _3 \ ] 4;©ó $ ØÙ%& ¤ #DDH 0~ë DIF¶ "C é *ð ^ °ê1Ô ¯ ~.03d.04D9 ¶ Ç 5 \ ] 4;ó $ ð ^ °¦ bAG Ç LM N~1U ?Q R S5
§ _2£_ 3º 1F J % cDIFDE constant<ASA DIF ¶ "C é *ÔW=W¼ Ç =~ T U ð K °ÂASAbE0.3
6Â_J 9 \ ?¶ @=À ± I £01S 0.3%& ¤ #DD H 0êÇ _üO P ÓÃ& ' ù M ð ß 5 ~+ ;¤ ã %& ¤ #~6ð
K °+ ;ÚI 5 \ ] 4;ó $ <¦A § 61c \ ] 4;+0
1w + ; , ¨©5~%& ¤ #LRT-ST#~DIF ¶ "C é *5A < _ü*LRT-SP#£LRT-PA#k l =DIFC é ÔWLRT-SP#ef üÍ gb+ ;5
_3 \ ] 4;©ó $ VDIFDE constant DIF% ASA
Type I error Power
L-ST L-SP L-PA L-ST L-SP L-PA 0% 0.00 .04 .03 .03 10% 0.04 .04 .03 .04 .93 .84 .89 20% 0.15 .04 .03 .04 .76 .71 .69 30% 0.19 .04 .04 .03 .73 .77 .72 40% 0.30 .04 .03 .03 .70 .71 .70 PDIFDE balanced _4 \ ] 4;ó $ ØÙ~DIFDE balanced<_J BU \ ] =À Ö ° ð ^ ~Wang2001 N Í 6<§ k Í ASA T é U æ 9 0V~ë & DIF¶ "C é *DH 01M N~U ?Q R DS5 =? º 1F J % 6ð K °§ k Í ASAL0%& ¤ #DD H 0º L?% ó cO P M N5 ~+ ;¤ ã LRT-ST#÷f ü³ + 3 4 Ù² u #£pure anchor#~6ð K °ÓÃ#i 0 ÚI 4 45
_5 \ ] 4;©ó $ ØÙ~ë & DIF¶ "C é *°6%& ¤ #DDH 0êÇ M NO P » ?% ?¶ @ 4;01D¨© t W¼
.015~+ ;¤ ã º ?% \ ] 4;01¨©%& ¤ #D+ ;L*z \ ] 4;ó $ <X A V~DIF ¶ "C é *" b<0n " ä 5=
DIF¶ "C é *W¼ %& ¤ #+ ;LqÔ Y Z ó 5D°LRT-ST
#_üg LRT-SP#£ LRT-PA#eÃ#f ü5 4 45
_ 4 \ ] 4;ó $ VDIFDE balanced DIF% ASA
Type I error Power
L-ST L-SP L-PA L-ST L-SP L-PA 0% 0.00 .02 .02 .02 10% 0.00 .02 .02 .02 .97 .98 .95 20% 0.00 .02 .03 .02 .99 .98 .97 30% 0.00 .03 .03 .03 .93 .92 .92 40% 0.00 .03 .03 .03 .93 .91 .89 _5 \ ] 4;©ó $ V DIFDE balanced DIF% ASA
Type I error Power
L-ST L-SP L-PA L-ST L-SP L-PA 0% 0.00 .03 .04 .04 10% 0.00 .04 .03 .04 .89 .86 .82 20% 0.00 .04 .03 .04 .91 .88 .88 30% 0.00 .04 .03 .03 .78 .79 .77 40% 0.00 .03 .03 .03 .78 .74 .70 %P Ø
m n [+é [%& ¤ #~& ð K D+B°ÂDIF¶ "C é *bE
40%PASAbE0.3DH 0L4M N~1U ?Q R SÓÃù M ð ß 5 ä | +=À DIF¶ "C é *ÔW%& ¤ #D+ ;LqÔ DY Z 5 =~ & ©' DIFDE °L1c \ ] 4;+01+ ;§ , I ¨©5 ~ DIFDE constantV \ ] 4;ó $ ð K °LRT-SP#£ LRT-PA#_üó 0©ä VDIF¶ "C é *ÔW ¤ #D+ ;1 ` b9 LRT-ST#5=~ \ ] 4;©ó $ ð ^ °1F J % c DIF¶ "
C é *© 9 30%<LRT-ST#_ü` ¼ 9 LRT-SP#£LRT-PA#=
DIF¶ "C é *W¼ d30%<LRT-SP#ef üÍ gb+ ;5 ~DIF DE balancedð K °©7 \ ] 4;ó $ £h LRT-ST#ù÷j ü ³ +3 4 ÙLRT-SP#£ LRT-PA#ÓÃ#i 0 ÚI u o 4 45a ] + =À ~ T U ð K °%& ¤ #~+ ; +01ó c. © º Â
_J %& ¤ #~GRM. °D DIF+3 4 4T é ó æ 5
%& ¤ #~©' DIF++3 ð ß 1á _6d_9¦Ù Ú5~& ©' ð K °c DIFð ^ £<%& ¤ #D+ ;L* z a ] + ;¦b=c DIFð ^ W¼ %<%& ¤ #D+ ;Âg. X A A£a ] + ;ØÙó 0©ä FÇ =~ù qDIF ð ^ °1F J % %& ¤ #D+ ;eÚI °X VA 9 a ] + ; Ñ 7 14Ö § 9
~¶ " +I £ ÷b6Â~6I £+¥ W¼ 0.3DIF ²DIF+3 4 Ùº , q½ ä ¨©5 _6 \ ] 4;ó $ V DIFDE constant1~4°D+ ; L-ST L-SP L-PA DIF% ASA 1 2 3 4 1 2 3 4 1 2 3 4 10% 0.04 .99 .99 1.00 1.00 .96 .98 20% 0.15 .99 .98 .90 .53 1.00 .96 .97 .52 1.00 .97 .93 .46 30% 0.19 .97 .95 .93 .48 .99 .97 .95 .58 .99 .97 .90 .55 40% 0.30 .98 .95 .91 .53 .98 .95 .97 .70 .99 .97 .94 .69 _ 7 \ ] 4;©ó $ VDIFDE constant1~4°D+ ; L-ST L-SP L-PA DIF% ASA 1 2 3 4 1 2 3 4 1 2 3 4 10% 0.04 .94 .92 .92 .75 .89 .88 20% 0.15 .97 .89 .70 .46 .90 .88 .62 .43 .97 .85 .62 .30 30% 0.19 .87 .80 .65 .39 .89 .85 .75 .41 .86 .80 .63 .35 40% 0.30 .89 .71 .76 .44 .90 .73 .77 .44 .89 .76 .73 .42
_8 \ ] 4;ó $ VDIFDE balanced1~4°D+ ; L-ST L-SP L-PA DIF% ASA 1 2 3 4 1 2 3 4 1 2 3 4 10% 0.00 .97 .98 .95 20% 0.00 .99 .98 .98 .98 .98 .97 30% 0.00 .97 1.00 .81 .99 .99 .78 .99 .98 .78 40% 0.00 1.00 .97 .99 .79 1.00 .98 .97 .71 1.00 .97 .96 .63 _9 \ ] 4;©ó $ VDIFDE balanced1~4°D+ ; L-ST L-SP L-PA DIF% ASA 1 2 3 4 1 2 3 4 1 2 3 4 10% 0.00 .89 .86 .82 20% 0.00 .96 .87 .90 .86 .90 .86 30% 0.00 .93 .86 .55 .93 .85 .61 .93 .83 .56 40% 0.00 .88 .80 .85 .58 .87 .78 .80 .53 .88 .79 .75 .38 á [°¶ " º 1ü Í DIF~£ð ^ °D+ ;0 153d5é > ÃDIFP~+ qDIF[\~+ q DIFDð K °§ G µ D ICC5 P0dP4ë Q _?¶ @~6¶ "+0é d 4é < 5*53£ 41F J Í DÚI <u 6Ñ 7 ~DIFð ^ °DIF Í < 1u b+ ;+vF ¤ ã " 3£51ü Í j ñ î a ] ³ P ! Óà ä z ¾ u <6Ñ 7 ~ DIF > + I 5H I + ;C Ç X A 5
3 ÃDIF¶ " 4 DIF¶ "
5 DIF¶ "
Ü £ ë B <= +3 ØÙDDH 0\+ ;§ , ¨ © %& ¤ #DØÙ <1é [Analysis of VarianceÓ[¿ ~ # Scheffé methed ú ´ *5[°é [ØÙé , DH 0£+ ; ã é Ù Ú5 DH 0 " _101ü Í ¨©DH 0D³ 6 7 <= G \ ] 4;0 F1,30=78.035Pη2=0.722[\¤ #0F2,30=5.462Pη2=0.2675" ØÙº I J ³ I N ò - ¨©G \ ] 4;0£DIFC é *F4,30=6.286P η2 η2
~%& ¤ #DH 0L?% \ ] 4;0¨©c \ ] 4;©ó $ <LRT-ST#DDH 0 q.01d.02W¼ À £LRT-SP#£LRT-PA#e t W¼ .01A§ 9 W¼ À £ó c© %& ¤ #DDH 0êM N~O P Q R S5 _ 10 DH 0<1é [ØÙ F Eta Abilitydifference 1 0.001 78.035 0.000 0.722 DIFpattern 1 < 0.001 0.347 0.560 0.011 Percentage 4 < 0.001 1.951 0.128 0.206 method 2 < 0.001 5.462 0.009 0.267 STSP abilitydifference * DIFpattern 1 < 0.001 0.000 1.000 0.000 abilitydifference * percentage 4 < 0.001 6.286 0.001 0.456 abilitydifference * method 2 < 0.001 0.780 0.467 0.049 DIFpattern * percentage 4 < 0.001 0.130 0.970 0.017 DIFpattern * method 2 < 0.001 3.728 0.036 0.199 percentage * method 8 < 0.001 0.585 0.782 0.135 30 < 0.001 60 a R ¤ = .990 (ô ´ R ¤ = .980) + ; " _111ü Í ¨©+ ;D³ 6 7 <= G \ ] 4;0 F1,23=525.271Pη2=0.958PDIFC é *F3,23=72.073Pη2=0.904PDIFDE F1,23=68.169Pη2=0.748[\¤ #0F2,23=4.717Pη2=0.2915" ØÙº I J ³ I N ò - ¨©G DIFDE £DIFC é *F3,23=39.191P η2=0.836P \ ] 4;0£DIFC é *F 3,23=7.707Pη2=0.501[\ \ ] 4 ;0£¤ #0F2,23=3.835Pη2=0.2505
] 4;§ ©ó $ Á ó $ <+ ;ÚI W¼ VcDIFC é *ÔW<+ ; W¼ À £º D¼ ä 5~LRT-ST#+ ; 1W¼ 14%DIFC é *W¼ d 1W¼ 20%F=LRT-SP#£LRT-PA#_üT é U æ c \ ] 4 ;ó $ <+ ; 1W¼ 21%d e1W¼ 26%£28%5 _11 + ;<1é [ØÙ F Eta abilitydifference 1 0.216 525.271 0.000 0.958
DIFpattern 1 0.028 68.169 0.000 0.748 balanced constant
percentage 3 0.030 72.073 0.000 0.904 10%20%30%40% method 2 0.002 4.717 0.019 0.291 STPA abilitydifference * DIFpattern 1 < 0.001 0.993 0.329 0.041 abilitydifference * percentage 3 0.003 7.707 0.001 0.501 abilitydifference * method 2 0.002 3.835 0.037 0.250 DIFpattern * percentage 3 0.016 39.191 0.000 0.836 DIFpattern * method 2 0.001 1.920 0.169 0.143 percentage * method 6 < 0.001 1.204 0.340 0.239 23 < 0.001 48 a R ¤ = 1.000 (ô ´ R ¤ = .999)
6 7 %& ©' ( ) *+ # ~GRM . Ó*5 LRT-ST
#PLRT-SP#[\ LRT-PA#~DIF+3 +DH 0\+ ;5 ØÙI J %& ¤ #DDH 0L \ ] ?¶ @ 4;D01=` vbAÂ~
\ ] 4;©' [\ASAbE 0.3ð ^ °G Ç 1M N~O P Q R S561Ù DH 0_ü£Woods2009( ) *+ # ~ $ , -. ï ð + ØÙ| 5 + ;¤ ã ~ & ©' DIFDE ?¶ @ 4;+01' È I ¨©%& ¤ #D+3 4 4 + ;L \ ] ?¶ @ 4;D01=Ú I °X 6i ü£Â 20109 é ï ð D ØÙó ' 5Ì § F J LRT-ST#PLRT-SP#[\LRT-PA#+3 ØÙD01ó c. © V% @D+ ;LqDIF¶ "C é *W¼ =°X Y Z 61Ù %& ¤ #~
DIF+3 +4 4T é U æ 5ä é [~DIFDE constantV \ ] 4;ó $ ð ^ °LRT-SP#D+ ;~ ð K ` b9 LRT-PA#= DIF¶ "C é *ÔW ¤ #D+ ;L1` b9 LRT-ST#F=c \ ] 4;©ó $ VDIF¶ "C é *© 9 30%<LRT-ST#_ü` ¼ 9 LRT-SP #£LRT-PA #Ç =DIF¶ "C é *W¼ d 30%<LRT-SP#ef üÍ gb+ ;5 ¤ ã cDIFDE balanced<©7 \ ] 4;ó $ £h ØÙi ü LRT-ST#+3 4 4` a 9 & ¤ #A01à © 5
m n [+é [1c ~GRM . °o %& ©' ( ) *+ # DIF +3 <DDH 06 7 ?% \ ] 4;01D¨©=+ ;¤ ã 6 7 e?
% \ ] 4;01PDIF¶ "C é *[\DIFDE D¨©5-@á T U o ³ ~ ] IRTLRDIF~GRM. °[( ) *+ #¦ DIF+3 i ü6 ] ~ ð K °÷Ï q +3 4 4Ç =~T ï ð \ ] ?¶ @~ 4;£ I £+01[\BU DIF¶ "C é * £LÉ @Ã#8M £r B6 6RS ~+3 ¼ ² u p DFTDq ` r s 1t u v+3 4 4W¼ +3 ØÙ w x 5
á IRTLRDIF( ) *+ # ~GRM . ØÙN Í LRT-ST#PLRT-SP#[\ LRT-PA#%@9 DIF¶ "+3 4 401©ä J I J Í IRTLRDIF ] ~ é "DIF +3 +T x 5=~´ +RS 1 u < ð K W¼ B <= 89 º p é [ ] $ [ Ò w á ( ) *+ #~GRM. °+3 4 45 ØÙº i ü~ GRM. °89 DIF ¶ "<DIF
¶ "+3 Í DIF < ÚI b9 DIF¶ "6ð ß I ¨©DIF¶ " a ] + ;56~ é . ¶ "I £01~ë I £+ ĺ 1, ¬ ¦ ¯ á ¤ 5 =~ Süq01¶ "2 4+3 ó y 8 6 "ê É9 é . ï ð 6RS ¬ ¦ @1 ² u [\DFTDq ` ~ é "01¶ "2 4+3 ¤ #+ ë é . ï ð +3 é [[ á ² u [\DFTDq ` ~ é "+3 ¤ #+DIF+3 4 45
 (2010)5 ; Å Æ ä Æ BU § k 7 8 ¬ Í Ç Å È 5
Andrich, D. (1978). A binomial latent trait model for the study of Likert-typed attitude questionnaires. British Journal of Mathematical and Statistical Psychology, 31, 84-98.
Ankenmann, R. D., Witt, E. A., & Dunbar, S. B. (1999). An investigation of the power of the likelihood ratio goodness-of-fit statistic in detecting differential item
functioning. Journal of Educational Measurement, 36, 277-300.
Bolt, D. M. (2002). A Monte Carlo comparison of parametric and nonparametric polytomous DIF detection methods. Applied Measurement in Education, 15, 113-141.
Candell, G. L., & Drasgow, F. (1988). An iterative procedure for linking metrics and assessing item bias in item response theory. Applied Psychological Measurement,
12, 253-260.
Cohen, A. S., Kim, S. H., & Baker, F. B. (1993). Detection of Differential Item Functioning in the Graded Response Model. Applied Psychological Measurement,
17, 335-350.
Cohen, A. S., Kim, S. H., & Wollack, J. A. (1996). An investigation of the likelihood ratio test for detection of differential item functioning. Applied Psychological
Measurement, 20, 15-26.
Clauser, B., Mazor, K., & Hambleton, R. K. (1993). The effects of purification of the matching criterion on the identification of DIF using the Mantel-Haenszel
procedure. Applied Measurement in Education, 6, 269-279.
Cheung, G. W., & Rensvold, R. B. (1999). Testing factorial invariance across groups: A reconceptualization and proposed new method. Journal of Management, 25, 1-27. Embretson, S. E., & Reise, S. P. (2000). Item response theory for psychologists.
Multivariate Applications Books Series. Mahwah, NJ, US: Lawrence Erlbaum
Associates Publishers.
Finch, H. (2005). The MIMIC model as a method for detecting DIF: Comparison with Mantel-Haenszel, SIBTEST, and the IRT likelihood ratio. Applied
PsychologicalMeasurement, 29, 278-295.
French, B. F., & Maller, S, J. (2007). Iterative Purification and Effect Size Use With Logistic Regression for Differential Item Functioning Detection. Educational and
Psychological Measurement, 67, 373-393.
Fidalgo, A. M., Mellenbergh, G. J. & Muñiz, J. (2000). Effects of amount of DIF, test length, and purification type on robustness and power of Mantel-Haenszel procedures. Methods of Psychological Research, 5, 43-53.
Hidalgo-Montesinos, M. D., & Gómez-Benito, J. (2003). Test purification and the evaluation of differential item functioning with multinomial logistic regression.
European Jouranl of Psychological Assessment, 19, 1-11.
Holland, P. W., & Thayer, D. T. (1988). Differential item performance and the Mantel-Haenszel procedure. In H. Holland & H. I. Braun (Eds.), Test validity (pp. 129-145). Hillsdale, NJ: Lawrence Erlbaum.
Kim, S.-H., & Cohen, A. S. (1998).Detection of differential item functioning under the graded response model with the likelihood ratio test. Applied Psychological
Hillsdale, NJ: Lawrence Erlbaum.
Mantel, N. (1963). Chi-square tests with one degree of freedom: Extensions of the Mantel-Haenszel procedure. Journal of the American Statistical Association,
58,690-700.
Masters, G. N. (1982). A Rasch model for partial credit scoring. Psychometrika, 47, 149-174.
Muraki, E. (1992). A generalized partial credit model: Application of an EM algorithm.
Applied Psychological Measurement, 16, 159-176.
Muraki, E. (1993). Information functions of the generalized partial credit model.
Applied Psychological Measurement, 17, 351-363.
Mantel, N., & Haenszel, W.(1959). Statistical aspects of the analysis of data from retrospective studies of disease. Journal of the National Cancer Institute, 22, 719-748.
Miller, M. D. & Oshima, T. C. (1992). Effect of sample size, number of biased items, and magnitude of bias on a two-stage item bias estimation method. Applied
Psychological Measurement, 16, 381-388.
Miller, T. R., & Spray, J. A. (1993). Logistic discrimination function analysis for DIF identification of polytomous scored items. Journal of Educational Measurement,
30 (2), 107-122.
Navas-Ara, M. J., & Gómez-Benito, J. (2002). Effects of ability scale purification on identification of DIF. European Jouranl of Psychological Assessment, 18, 9-15. Park, D. G., & Lautenschlager, G. J. (1990). Improving IRT item bias detection with
iterative linking and ability scale purification. Applied Psychological
Measurement, 14, 163-173.
Mantel-Haenszel procedures for detecting differential item functioning. Applied
Psychological Measurement, 17, 105-116.
Reise, S.P., & Yu, J. (1990). Parameter recovery in the graded response model using MULTILOG. Journal of Educational Measurement, 27, 133-144.
Raju, N, S., van der Linden, W., Fleer, P. (1995). An IRT-based internal measure of test bias with applications for differential item functioning. Applied Psychological
Measurement, 19, 353-368.
Samejima, F. (1969). Estimation of a latent ability using a response pattern of graded scores. Psychometrika Monograph Supplement, 17,1-100.
Stark, S., Chernyshenko, Oleksandr, S., & Drasgow, F. (2006). Detecting differential item functioning with confirmatory factor analysis and item response theory: Toward a unified strategy. Journal of Applied Psychology, 91, 1292-1306.
Swaminathan, H. & Rogers, H. J. (1990). Detecting differential item functioning using logistic regression procedures. Journal of Educational Measurement, 27, 361-370.
Shealy, R., & Stout, W. (1993). A model-based standardization approach that separates true bias/DIF from group ability differences and detects test bias/DTF as well as item bias/DIF. Psychometrika, 58,159-194.
Su, Y.-H. & Wang W.-C. (2005). Efficiency of the Mantel, generalized
Mantel-Haenszel, and logistic discriminant function analysis methods in detecting differential item functioning for polytomous items. Applied Measurement in
Education, 18(4), 313-350.
Shih, C.-L. & Wang W.-C. (2009). Differential Item Functioning Detection Using the Multiple Indicators, Multiple Causes Method with a Pure Short Anchor. Applied
Thissen, D., Steinberg, L., & Gerrand, M. (1986). Beyond group mean differences: The concept of item bias. Psychological Bulletin, 99, 118-128.
Thissen, D., Steinberg, L., & Wainer, H. (1988).Use of item response theory in the study of group differences in trace lines. In H. Wainer & H. I. Braun (Eds.), Test
validity (pp. 147-169). Hillsdale NJ: Lawrence Erlbaum.
Thissen, D., Steinberg, L., & Wainer, H. (1993). Detection of differential item
functioning using the parameters of item response models. In P. W. Holland & H. Wainer (Eds.), Differential item functioning (pp. 67-113). Hillsdale NJ: Erlbaum. Wang, W.-C. (2001, September). Effects of anchor item methods on the detection of
differential item functioning within the family of Rasch models. Paper presented at
the annual meeting of the Chinese Psychological Association, Chia-Yi, Taiwan. Manuscript submitted for publication.
Wang, W.-C. (2004). Effects of anchor item methods on differential item functioning detection within the family of Rasch models. Journal of Experimental Education,
72, 221-261.
Wang, W.-C. (2008). Assessment of differential item functioning. Journal of Applied
Measurement, 9, 387-408.
Woods, C. M. (2009). Empirical selection of anchors for tests of differential item functioning. Applied Psychological Measurement, 33, 42-57.
Wang, W.-C., & Yeh, Y.-L. (2003). Effects of anchor item methods on differential item functioning detection with the likelihood ratio test. Applied Psychological
Measurement, 27, 479-498.
Wang, W.-C. & Su, Y.-H. (2004a). Effects of average signed area between two item characteristic curves and test purification procedures on the DIF detection via the Mantel-Haenszel method. Applied Measurement in Education, 17, 113-144.
Wang, W.-C., & Su, Y.-H. (2004b). Factors influencing the Mantel and generalized Mantel-Haenszel methods for the assessment of differential item functioning in polytomous items. Applied Psychological Measurement, 28(6), 450-480.
Wang, W.-C., & Shih, C.-L. (2010). A new strategy to assess differential item functioning. Educational and Psychological Measurement.
Wang, W.-C., Sun, G.-W., & Shih, C.-L. (2010). A new strategy to assess differential item functioning. Paper submitted to Educational and Psychological
Measurement for publication.
Zumbo, B. D. (1999). A handbook on the theory and methods of differential item
functioning(DIF): Logistic regression modeling as a unitary framework for binary and Likert-type (or ordinal) item scores. Ottawa, Canada: Directorate of
Human Resources Research and Evaluayion, Department of National Defense. Retrieved from http://www.edu.ubc.ca/faculty/zumbo/DIF/index.html.
Zwick, R., Donoghue, J. R., & Grima, A. (1993). Assessment of differential item functioning for performance tasks. Journal of Educational Measurement, 30, 233-251.