The purpose of this study was to develop criteria fordetecting sex biased items from empirical data for fourstatistical indices

signed area (SA) and unsigned area (UA)based on item response theory, 2-parameter model, MH fromMantel-Haenszel procedure and ('B)SSIB from SIBTEST, using theitem response data to English and mathematic entrance examinationtests to higher education, administered by the Ministry of theUniversity Affairs in 1992. The criteria identifying biased itemsfor each index were developed by calculating arithmetic means ofthe index values from 100 pairs of examinees of the same sex ineach condition of 7 test lengths

20 30 and 40 items formathematics and 50 60 70 and 80 items for English, and 6 sizes ofexaminees

100 200 400 600 800 and 1,000. Two sets of criteria of each index were set based on themeans, with and without regarding to the differences across thetest lengths and the sizes of examinees. The major findings wereas follows: 1. The criteria developed from the empirical data to detectsex biased items were..etc.