Abstract:
So far there have been Studies the consistency of the difficulty parameters (bs) of the test items analyzed by means of Rasch Model only when the samples were from the same populations or were indepdent and normally distributed. Therefore, the main purpose of this study was to investigate the consistency of such parameters when the samples were from 3 different populations with different learning ability levels, the same kurtosis and equal variances : 1) positive skewness and low ability, 2) normal distribution and moderate ability, and 3) negative skewness and high ability. A Monte Carlo Simulation technique was applied by means of using computer simulations. The so-called subjects by of each group were 500 and supposed to take a 30 item test in 100 simulations. The findings were summarized as follows : 1. The estimated difficulty parameters (bs) for moderate items were consistent. From 30 items, 19 items from the low and high ability groups, and 22 items from the moderate ability, were found in such a circumstance. 2. The standard errors of estimated parameters showed nonsymmetrically distributed for which examinees in low and high ability. Howevers, for the moderate ability group, the standard errors of estimated were symmetrically distributed. 3. The standard errors of the difficulty indices of the 3 group of examinees were not significantly different (P< .05)