@article{oai:nagoya.repo.nii.ac.jp:00002447,
 author = {野口, 裕之 and NOGUCHI, Hiroyuki},
 journal = {名古屋大學教育學部紀要. 教育心理学科},
 month = {Dec},
 note = {The purpose of this paper is to study how the true item difficulty parameter distribution in a test affects the item parameter estimates by heuristic method. Two types of tests were constructed. The one is "uniformly type test" in which the true item difficulty parameters have uniformly distribution from easy to hard. The other is "trancated type test" in which the true item difficulty parameters have negatively skewed distribution. Test length were fixed at 20,40 and 80 items and the number of subjects were fixed at 1000,2000 and 4000. The three factors-test length (3 levels), difficulty parameter distribution (2 levels) and number of subjects (3 levels)-were completely crossed to simulate 18 testing situations. The subjects' abilities were randomly sampled from the unit normal distribution. The item response data of each subject were generated using the uniform random number generator in computer and the item characteristic functions. The item parameters were separately estimated by heuristic method for all eighteen data sets. The estimated item parameters were compared to the true item parameters in all data sets. It was concluded that heuristic method practically lead to good item parameter estimates, if the skewness of true item difficulty distribution in a test was not extreme but moderate. The number of items and/or subjects rather than true item difficulty parameter distribution affect the estimates of item parameters., 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。},
 pages = {15--21},
 title = {<原著>項目困難度の分布の偏りが IRT 項目パラメタの発見的推定値に与える影響},
 volume = {39},
 year = {1992}
}