Akademik Veri Yönetim Sistemi
JOURNAL OF MEASUREMENT AND EVALUATION IN EDUCATION AND PSYCHOLOGY-EPOD, cilt.6, sa.2, ss.331-350, 2015 (ESCI)
Under the scope of this study; parameter estimations of simple and complex structured two and three dimensional tests have been performed according to different test length and sample size conditions; according to the estimations, RMSE, bias and relation between estimated and actual parameters have been investigated. Research data has been generated by means of simulations according to requirements of two dimensional simple structure, two dimensional complex structure, three dimensional complex structure and three dimensional complex structure. The simulation conditions were test length (12 and 48 items) and sample size (1000, 2000 and 4,000). By intercrossing the simulation conditions (4 test structures x 2 test lengths x 3 sample sizes); 24 crossed conditions have been acquired. By fixing item and ability parameters and applying 25 replications; 600 data sets have been generated and analyzed. According to the results, parameter recovery of item parameters increases with the number of items and examinees. For each condition; d parameters have been estimated more accurately than a parameters on the same pattern. Parameter recovery of ability parameters is developed with the increase in test length. High correlation values acquired in the conditions where RMSE values are low. Additionally; suggestions on the required test length and sample size for accurate estimations are provided at the final stage of the study.