공통문항유형과 검사 차원성에 따른 혼합형 검사의 IRT 동등화 = Comparison of scale linking results in mixed-format test-under the different conditions of common item format and test dimensionality

다국어 초록 (Multilingual Abstract)

In scale linking of mixed-format test, if an item type has more information than the other and the common items are made up with only a single format, the common items may not represent the entire test in terms of item format. Considering that they should be a miniature of the whole test, common items should still have representativeness in item format. In multidimensional situation, scale linking through only single item type takes no account of multidimensionality and can affect the linking results. The major purpose of present study is to examine scale linking results under the conditions of different combination of common item format and test dimensionality using IRT linking methods.
Three factors were investigated for the simulation study. 3 types of test dimension, 3 types of mixed-format test and 4 types of common item set and the combination of the these factors led to 39 conditions which were compared under the four linking methods.
As a results, under the unidimensional situation, the test where one item type has more information than the other led to lower MSE when linking through relatively more informative common items. On the other hands, when there was the presence of high multidimensionality, the balanced test type in the amount of the information showed lower MSE when linking through mixed-format common items. Moreover when mixed common item sets represented the item format composition of the entire test, MSE values tended to decrease. This means that if a mixed format test is multidimensional with respect to item format the accuracy of unidimensional scale linking through single-format common items could not be guaranteed. And balanced test type where item formats are equivalent in the amount of the information leads to lower linking error and it is desirable for the common item sets to be mixed format rather than a single format and to represent the item formats of entire test.

국문 초록 (Abstract)

혼합형 검사가 다차원성을 지닐 때 동등화에 있어서 단일 문항유형으로 구성된 공통문항은 문항유형에 대한 대표성을 갖기 어려우며 동등화 결과에도 부정적인 영향을 미칠 수 있다. 본 연구는 혼합검사유형, 공통문항유형과 검사 차원을 탐색요인으로 하여 문항반응이론 동등화 방법에 의한 결과를 비교하였다. 다분문항과 이분문항이 제공할 수 있는 정보(문항반응이론의 기대정보함수)의 양에 근거하여 혼합형 검사와 공통문항을 각각 세 유형과 네 유형으로 구분하고 각 문항유형이 측정하는 구인의 유사성에 따라 검사차원을 세 조건으로 분류하였다. 네 가지 문항반응이론의 동등화 방법으로 39개 조건에서의 혼합형 동등화를 각각 실행하였고 두 가지 준거에 근거하여 동등화 결과를 평가, 비교하였다. 연구결과, 검사 차원성에 따라 혼합형 검사유형과 공통문항유형은 다른 양상을 보였다. 일차원의 경우, 특정문항유형이 상대적으로 많은 정보를 제공하는 검사에서는 공통문항유형도 정보를 많이 제공하는 문항유형으로 구성될 때 동등화 오차가 감소하였다. 반면 다차원의 경우 차원 간 상관이 낮을수록 각 문항유형별로 제공하는 정보의 양이 유사한 검사유형에서 안정적인 동등화 결과를 보였으며 공통문항유형은 혼합유형일 때, 특히 전체 검사의 문항유형 구성을 대표할 때 동등화 오차가 감소하는 경향을 보였다.

목차 (Table of Contents)

• 요약
• Ⅰ. 서론
• Ⅱ. 문항반응이론 동등화
• Ⅲ. 연구방법
• Ⅳ. 결과
• Ⅴ. 결론 및 논의
• 참고문헌
• ABSTRACT
• 부록

참고문헌 (Reference)

1 "Test equating,Scaling, and Linking" Springer-Verlag 2004

2 "Test equating under the multiple-choice model" 26 : 255-270, 2002

3 "Stable response functions with unstable item parameter estimates" 26 : 239-254, 2002

4 "Some latent trait models and their use in inferring an examinee's ability Statistical Theories of Mental Test Scores" 397-472, 1968

5 "Performance of a proposed method for the linking of mixed format tests with constructed response and multiple choice items." 37 : 329-346, 2000

6 "Paper presented at the Annual Meeting of the National Council on Measurement in Education." 2002

7 "PARSCALE (version 4.1) IRT item analysis and test scoring for rating- scale data." Scientific Software, Inc. -4, 2003

8 "Optimally defining criterion functions for the Characteristic curve procedures in the IRT scale linking" 2004

9 "On the equivalence of the traits assessed by multiple-choice and constructed-response test Construction versus choice in cognitive measurement" Lawrence Erlbaum Associates 29-44, 1993

10 "Obtaining a common scale for item response theory item parameters using separate versus concurrent estimation in the common-item equating design" 26 : 3-24, 2002

11 "Numerical methods for unconstrained optimization and nonlinear equations." Society for Industrial and Applied Mathematics. 1996

12 "Least squares estimation of IRT scale linking coefficients under the graded response model" 2004

13 "LISMIX [computer program]" Korea Institute of curriculum and evaluation. 2004

14 "Journal of Educational Measurement" 179-193, 1980

15 "Item characteristic curve solution to three intractable testing problems" 14 : 139-160, 1977

16 "IRT scaling linking methods for Mixed-format tests" 2004

17 "Factor analysis for items or testlets scored in more than two categories" Lawrence Erlbaum Associates 217-250, 2001

18 "Estimation of a latent ability using a reponse pattern of graded scores" 1969

19 "Estimating item parameters and latent ability when responses are scored in two or more nominal categories" 37 : 29-51, 1972

20 "Equivalence of free-response and multiple-choice items Journal of Educational Measurement" 77-92, 1991

21 "Equatinjg tests under the nominal response model" 17 : 239-251, 1993

22 "Equating tests under the graded response model" 16 : 87-96, 1992

23 "Equating logistic ability scales by a weighted least squares method" 22 : 144-149, 1980

24 "Empirical sampling distributions of equating coefficients for graded and nominal response instruments" 21 : 157-172, 1997

25 "Developing a common metric in item response theory" 207-120, 1983

26 "Are tests comprising both multiple-choice and free-response items necessarily less unidimemsional than multiple-choice tests? An analysis of two tests Journal of Educational Measurement" 113-123, 1994

27 "Applications of item response theory to practical testing problems" Erlbaum 1980

28 "Application of an EM algorithm" 373-383, 1992

29 "An investigation of linking methods under the graded response model" 22 : 116-130, 1998

30 "An empirical examination of the IRT information of polytomous Scored reading items under the generalized partial credit model Journal of Educational Measurement" 295-311, 1992

31 "A rating formulation for ordered response categories" 43 : 561-573, 1978

32 "A minimum ┶2 method for equating tests under the graded response model" 19 : 167-176, 1995

33 "A general model for free response data" 1972

34 "A Rasch model for partial credit scoring" 47 : 149-174, 1982