A Methodological Study on the Equivalence of Translation and the Differential Item Functioning for Crossnational Comparative Surveys

2005 Fiscal Year Final Research Report Summary

• Project/Area Number: 15530399
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: Social psychology
• Research Institution: Nagoya University
• Principal Investigator: NOGUCHI Hiroyuki Nagoya University, Graduate School of Education and Human Developmental Sciences, Professor, 大学院教育発達科学研究科, 教授 (60114815)
• Co-Investigator(Kenkyū-buntansha): FUJIMOTO Akira Shizuoka University, Law School, Professor, 大学院法務研究科, 助教授 (80300474)
• Project Period (FY): 2003 – 2005
• Keywords: DIF / Item Response Theory / International Comparative Study / BILOG-MG / Psychological measurement Scale / equivalence of translation

Research Abstract

1.In this study, we examined the. applicability of the Item Response Theory, which the test theory has put to a good use, to the detection of DIF(Differential Item Functioning) within the items of some measurement scales in crossnational comparative surveys. Also we examined how each DIF detection method showed the different results.
2.The item equivalence in crossnational comparison implies both "equivalence of translation" and "equivalence of meanings." The DIF analysis concerns the latter, though.
3.Among the various DIF analysis methods, we pick the following four. Two parametric methods(1.DIF command of BILOG-MG, 2.Thissen's likelihood ratio test) and two non-parametric methods(3.M-H statistics, 4.logistic regression.)
4.We applied these methods to the following data. 1.Legal Image Scale from a Japan-US law student survey. 2.The Scale for the Level of Confidence in various governmental institutions from a Japan-US-China legal consciousness survey. 3.F-Scale from a 22-country/region l aw and business student survey for contract consciousness.
5.Results. 1.Even in the case that the scale has. higher unidimensionality in factor analysis and confirmed equivalence of translation, we found some item shows DIF. 2.In the case that the number of samples are quite large, any statistical significance testing is likely to have significant results, we need to pay much attention to the possible different meanings of the terms used in the items. For example, 'Fairness' in the confidence in government scale. 3.We also found it is useful to use multi DIF detection methods to detect DIF items.
6.We presented our findings at the following academic conferences and meetings. 1.The Fifth Annual Conference of the Japanese Society for Law and Psychology(October 16.17,2004.) 2.The Sixth Annual Conference of the Japanese Society for Law and Psychology(October 15.16,2005.) 3.A meeting of the Asian-Pacific Dispute Resolution Project at the Institution of Asian Research, University of British Columbia(March 5- 11,2006.)