أثر عدد البدائل وطريقة التصحيح في الخصائص السيكومترية للاختبارات التحصيلية
This study aimed to identify the effect of the alternatives number and scoring method on the psychometric properties of multiple choice test (validity, and reliability), as well as in the statistical distribution properties, the study also aimed to determine the optimal number of alternatives for multiple choice tests and the best method to correct multiple-choice tests. To achieve the aims of this study, the researcher built achievement test with standard reference of multiple-choice test type in the following units: third unit (Engineering), fourth unit (Fractions), and the fifth unit (Measurement) of math to sixth grade of elementary education in the second term of the academic year 2012/2013 m. The test consisted of 40 items of multiple choice test type. Each item had five alternatives. According to the results of exploratory experimentation, the researcher built three models of the test. The three models are similar in stem, and equal in the number of items (32 items) for each model, but different in the number of alternatives as follows: 1) The first model: each item contained five alternatives. 2) The second model: each item contained of four alternatives, by discarding the least effective distractor from each item of the first model. 3) The third model: each item contained of three alternatives, by discarding the least effective distractor from each item of the second model. The test was also put the test instructions for each model correction of the four following methods: 1) The Number- Correct Method. 2) Correction for Guessing Methods “Rewarding”. 3) Correction for Guessing Methods “Reward and Punishment”. 4) Subset Selection Method. The study sample consisted of 1200 pupils from sixth grade of elementary education, who had studied in public schools in Latakia in the academic year 2012/2013 m. The sample was divided into four main groups representing the four methods of scoring, each group consisted of 300 pupils per 100 for each model, thus, the researcher had 12 subset distributed groups on 12 deliberately selected schools in Latakia. The researcher followed experimental and descriptive analytical approaches. The most important results of the study were: 1) There were 110 statistically significant differences at the level of significance 0.05 between the mean of reliability coefficients of the achievement test degrees of a multiple choice test, that had calculated by: (alpha Cronbach, split half according to Spearman- Brown, Guttman) in mathematics subject of sixth grade of elementary education attributed to the difference in scoring method or alternatives number. 2) There were statistically significant differences at the level of significance 0.05 between the mean of validity coefficients of the achievement test decrees of a multiple-choice test, that had calculated by (concurront, predictive) in math subject of the sixth grade of elementary education attributed to the difference in scoring method or alternatives number. 3) For the characteristics of the statistical distribution has been shown that the highest value of the test with three models were for mean of pupils degrees on the Subset Selection Method. The results also showed that the variation coefficient values that pupils degrees were more identified and stable in the model (3 alternatives) that corrected by Subset Selection Method. In the model (4 alternatives) that was corrected by The Number- Correct Method and in the model (5 alternatives) that was corrected by Correction for Guessing Methods “Reward and Punishment”. The results also showed that torsion coefficient values in the model (3 and 4 alternatives) that were corrected by The Number- Correct Method were closer to the normal distribution. 4) There were statistically significant differences at the level of significance 0.05 between the mean of the students degrees in the achievement test with (five, four, three) alternatives in math subject to the sixth grade of elementary education attributed to the difference in scoring method. Where the Subset Selection Method was the best method, next Correction for Guessing Methods “Reward and Punishment”, next Correction for Guessing Methods "Rewarding”, then The Number- Correct Method. 5) There are statistically significant differences at the level of significance 0.05 between the mean of the students degrees in the achievement test that was corrected by (The Number- Correct Method, Correction for Guessing Methods “Rewarding”, Correction for Guessing Methods “Reward and Punishment”, Subset Selection Method) in math subject for the sixth grade of elementary education attributed to the difference in the number of alternatives. Where the model of the three alternatives was the best, next the model of four alternatives, then the model of five alternatives. (Author’s abstract)