Page No. 344-353
Satyendra Nath Chakrabartty* & Rumki Gupta**
Former Director, Indian Maritime University,Kolkata Campus* & Indian Statistical Institute
Kolkata**
About 1147 school students were selected based on a pre-designed sampling scheme
and they underwent a Bullying test where each item had 5-response categories. Class
teachers independently marked the students as bully or non-bully by giving a holistic
rating in a qualitative fashion without any knowledge of students’ test scores. Validity of
the bullying scale was found by various approaches like Point bi-serial correlation, Chisquare measure of association and function of chi-square like Contingency Coefficient,
Phi Coefficient, R2 of Logistic regression separately for 2, 3, 4 and 5 response categories
obtained by a collapsing data and Hit ratios. Fluctuations of validity indices were
investigated to find the number of response categories, which performs best. Goodness
of classification of students was done in two groups namely, bully and non-bully owing
to the results of test scores and teachers’ ratings. The two-point, three-point and fourpoint scales performed rather poorly and indices were higher for scales with 5-response
categories. However, factor analysis showed that only one factor was measured by all
the response categories implying unidimensionality of the Bullying Scale even if data
are collapsed to have various response categories..