Abstract:
The purposes of this study were to find out any effects of four different instructions and types of scorings, viz. 0-1 method, Coombs' method, Anant's method and researcher's method, on the reliability, validity and discriminating power of multiple-choice tests. The subjects were 240 Mathayom Suksa Four students 138 boys and 102 girls -- of Surasakmontree School and Wat Noinai School. All of them took mathematics 011. The subjects were devided into four groups with equivalent mathematics achievement scores in first semester. The data were collected and the analysis were done . The reliability of each group was evaluated by Hoyt's anaIysis of variance ; the concurrent validity by Pearson s product moment correlation coefficient; and the discriminating power by the biserial correlation formula. To make the testing hypothesis possible, the Fisher's Z coefficient was transformed to compare the reliability, concurrent validity and discriminating power of the test resulting from the four groups. The results were as follows: 1. The reliability of the test based on the researchers instruction and type of scoring showed higher value than others. The reliability of the test based on Coombs instruction and type of scoring was higher than that one based on of 0-1 method. (P < .05) 2. The validity of the test based on the researchers instruction and type of scoring was higher than that one based on of 0-1 method. (P <.01) The validity of the test based on Anants instruction and type of scoring was higher than one based on of 0-1 method. (P < .05) 3. The difference of discriminating power of the four instructions and types of scorings was found no statistical significance.