The Effect of IRT Scale Transformation Procedures on Equating

국문 초록 (Abstract)

  IRT를 이용한 검사동등화에서는 동등화가 이루어지기 전에 동등화에 쓰여 지는 동형검사들의 문항모수들이 같은 능력척도에 놓여야만 한다. 능력척도를 변환시켜주는 여러 방법들이 존재하지만, 어떤 방법이 가장 정확하게 척도변환을 이루며, 결과적으로 가장 오차가 작은 동등화를 성취하는지 다양한 조건아래 검사해 보아야한다. 이 연구는 시뮬레이션을 통하여 서로 다른 사례수와 문항 개수를 가진 데이터를 생성한 후 5가지 척도변환방법을 비교하였다. 5가지 방법에는 동형검사들의 문항모수를 개별적으로 추정(separate estimation) 한 후 mean/mean, mean/sigma, Haebara 및 Stocking-Lord 방법을 사용하여 능력척도를 변경시키는 절차와 동형검사들의 문항모수를 동시추정(concurrent estimation)함으로써 자동적으로 척도변환을 이루는 절차가 포함되었다. 연구결과, 일반적으로는 척도변환을 잘 성취한 방법이 동등화도 정확하게 이루어내는 경향이 있었지만, 예외적인 경우도 발생하였다. 척도변환방법들은 사례수, 문항 개수, 그룹 간 능력 차이에 따라 다른 성취도를 보였다.

다국어 초록 (Multilingual Abstract)

  Application of IRT often requires that the item parameter estimates for test forms administered to different groups in different time should be placed on a common ability scale. There are many possible procedures for obtaining an IRT common scale. This study aims to provide measurement practitioners with guidance in selecting a procedure that is most suitable to their particular situations. This study implemented a series of simulations under a random groups equating design to compare the performance of five scale transformation procedures for obtaining a common scale and achieving less equating error: concurrent estimation, and four linking methods associated with separate estimation including the mean/mean, mean/sigma, Haebara, and Stocking-Lord methods. The results showed that the performance of five scale transformation procedures varied depending on sample size, test length, and level of group difference.

목차 (Table of Contents)

• Ⅰ. Introduction
  Ⅱ. IRT Linking Methods
  Ⅲ. Results
  Ⅳ. Conclusion and Discussion
  References
  저자소개
  〈요약〉
  

참고문헌 (Reference)

1 "Test equating, scaling, and linking: methods and practices. (2nd ed.)" New York: Springer-Verlag. 2004

2 "ST: A computer program for IRT scale transformation, version 1.0" Iowa City, IA: ACT 1995a

3 "PIE: A computer program for IRT equating, version 1.0" Iowa City, IA: ACT 1995b

4 "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

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

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

7 "IRT command language" Montrey, CA: Author 2001

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

9 "Developing a common metric in item response theory" 201-210, 1983

10 "Comparison of three IRT linking procedures in the random groups equating design" Iowa City, IA: University of Iowa 23 : 2007

11 "BILOG-MG: Multiple-group IRT analysis and test maintenance for binary items" Chicago: Scientific Software International 1996

12 "A study on IRT fixed parameter calibration methods using BILOG-MG" 19 (19): 323-342, 2006

13 "A comparison of linking and concurrent calibration under the graded response model" 26 : 25-41, 2002