Contrasting Group Method

Procedure

Before administration of the test, class-teachers of the selected schools were requested to mark each student selected from the class as Bully or Non-Bully. The teachers were well acquainted with their student’s behaviors (at least class-room behaviors) and with rating guides which was prepared beforehand. No student was evaluated by more than one teacher. Thus, classification of students by teachers was done by holistic rating in a qualitative fashion without any knowledge of students’ test scores. The situation is similar to Contrasting Groups Method as described by (Livingston & Zieky, 1982) which requires panelists to classify examinees according to whether they have or have not met the performance standard. Here, holistic rating

Higher value of mean test score for those identified as Bully by teachers in comparison to those identified as Non-bully for each Response category tends to imply face validity of the test. t -ratios for testing equality of means and F- ratios for testing equality of variances between bully and non-bully groups for different response categories are given in Table – 2

Observations

i) Mean test score for the combined group declined steadily with reduction in number of response categories. Difference of mean scores between i-th and (i -1) th response categories (for i =5, 4, 3) were found to be significant. Similar trends were observed for the Bully group as well as for the Non-bully group. Mean for Bully group was more than the same for Non-Bully group for each response category. ii) Variance of test score for the combined group decreased in a non-linear fashion with decrease in number of response categories indicating reduction in number of response categories tended to increase group homogeneity as can be seen from the Figure -

However, extent of homogeneity achieved was not found to be linearly related with number of response categories. Similar trend was observed for the Bully group as well as for the Non-bully group. However, null hypothesis of equality of variance of bully and non-bully groups could not be rejected for each response categories. iv) t -ratio for testing equality of means of Bully and Non-bully groups was found to be significant at 99% level only for 4-response