Comparing Ability Estimation Accuracy Using Test Length and Ability Parameter Distribution Based on Three Parameter Logistic Model (3PLM) Using Real and Generated Test Scores

Authors

AMER MAHMOUD AL-SOBUH M.A, Jordanian Ministry of EducationFollow

Abstract

The study aimed to compare ability estimation accuracy using test length and ability parameter distribution based on Three Parameter Logistic Model (3PL) using real and generated test scores. to answer the question, which is about the validity of the use of the data generated in the examination and comparison between the statistics and models provided by item response theory , and to achieve the aim of the study, eighth grade national science scores for the school year 2012\2013 were used as the real scores in the study and generate scores were obtained using the real scores for the same sample via WINGEN 3.0 software. Examinees' abilities and the estimation standard error were computed via maximum likelihood and expected a posterior used in BILOG-MG 3.0 software for the real and generated samples. After the suitable statistical analysis procedures were performed (paired sample t-test and effect size statistics for the dependent sample t-test), it was found that there is no practical differences in the estimation accuracy for the ability parameter based on the nature of scores (real, generated) based on the test length (25 and 50) items, and the distribution of the ability parameter (normal distribution , positively skewed, negatively skewed).

Recommended Citation

AL-SOBUH, AMER MAHMOUD M.A (2021) "Comparing Ability Estimation Accuracy Using Test Length and Ability Parameter Distribution Based on Three Parameter Logistic Model (3PLM) Using Real and Generated Test Scores," Jerash for Research and Studies Journal مجلة جرش للبحوث والدراسات: Vol. 22: Iss. 2, Article 22.
Available at: https://digitalcommons.aaru.edu.jo/jpu/vol22/iss2/22