الملخص الإنجليزي
This study aimed to examine the effectiveness of person fit statistic (Wright & Master weighted, Drasgow and Almehrizi weighted) in individual response models when the strength of local dependence differs between items (0.0, 0.3, 0.6, 0.9) and types of the model parameters by using generated data for a test contains 60 items. The values of the discriminty parameter ranged in the period (0.19 , 1.79] and the values of difficulty parameter ranged between -2 to +2. Apility was generated for 10000 persons, so that a normal distribution is distributed with a mean of zero and standard deviation of one.
The program (multilog 7.03) was used to estimate parameters of the items and the abilities of individuals. Also appropriate statistical methods were used to check Response Theory assumptions for dual parameters model. In addition, means, standard deviations, Skewness and Kurtosis were calculate to find repetitive distribution properties for each indicator and calculate the number of people who do not match each indicator of study indicators.
The results of the study indicated an increase in the difference between the number of non-conforming individuals in real data and the estimated increase in local dependence between items. It departs from the reference situation (local independent). The largest coefficient of agreement was between Wright indicator and Drasgow indicator. It reached 0.93 in local independent situation. The agreement increased in cases (0.3, 0.6, 0.9) for local dependence.
Moreover, the results of the study showed that the differences in the percentage of non-conforming individual in real and estimated data, are a statistical function for all three indicators in all cases at the level of significance 0.05. The percentage of non-conforming individual in real data in indicators is higher than the percentage of non-conforming individual in estimated data in all local dependence cases.
