2022-11-28T15:29:58Zhttps://nagoya.repo.nii.ac.jp/oaioai:nagoya.repo.nii.ac.jp:000024042021-03-01T21:20:59Z<ORIGINAL ARTICLE>Marginal Maximum Likelihood Estimation of the Equating Coefficients for Two IRT Scales Using Common Subjects' Design<原著>共通被験者デザインにおける等化係数の周辺最尤法による推定野口, 裕之6625NOGUCHI, Hiroyuki G.6626item response theoryequatingitem parametersmarginal maximum likelihood estimationcommon subjects' designIn the present paper, an equating method of two scales upon two separate tests using common subjects' item response patterns for both tests is proposed, and examined its applicability in actual IRT scales' equating situations. We assume that there are two IRT scales based upon test X and Y respectively. In the new method, each subjects' item response patterns for various tests are obtained, and then marginal maximum likelihood estimates of the equating coefficients k, l are extracted from these data, while, simultaneous maximum likelihood estimates were extracted in Noguchi (1986). The equating coefficients k, l are estimated by solving simultaneous non-linear equations, (29) and (30). In this paper, test X and Y are formed with 10 ideal test items respectively. Item parameters of them are given in Table 1. By computer simulational item response data for these tests, validity of this method is examined in two situations, such as changing the true equating coefficients (k=1.0,l=0.0 and k=0.9,l=-0.8) and the number of subjects (N=250,500,1000,2000). As a consequence, this new equating method proves to be valid in almost all cases in the simulational studies.国立情報学研究所で電子化したコンテンツを使用している。departmental bulletin paper名古屋大学教育学部1990-12-25application/pdf名古屋大學教育學部紀要. 教育心理学科37191198http://hdl.handle.net/2237/382703874796https://nagoya.repo.nii.ac.jp/record/2404/files/KJ00000137340.pdfjpn