Comparison of Computerized Classification Tests under Item Response Theory and under Stochastic Curtailment for Three-Category Classification.: Shortening the SAQI for assessing low, moderate and high-risk behavior

In educational research, the length of questionnaires has to be considered by test developers to ensure minimal respondent burden. Efficient screeners, like computerized tests, have the potential to shorten tests without losing information. In this paper, two computerized testing methods for classification into categories are discussed: Computerized Classification Test (CCT) under Item response theory (IRT) and CCT under Stochastic Curtailment (SC). In a post-hoc simulation study, using the item scores on the School Attitude Questionnaire Internet (SAQI) of a large sample of Flemish students (N = 4211), the CCT methods were compared with respect to respondent burden, given minimized differences in proportion of correct classifications. Both CCT under IRT and CCT under SC showed the potential to reduce the respondent burden while maintaining 90% concordance with the full-length questionnaire. CCT under SC required fewer items (20% of the total item set) than CCT under IRT (37%), but...Show moreIn educational research, the length of questionnaires has to be considered by test developers to ensure minimal respondent burden. Efficient screeners, like computerized tests, have the potential to shorten tests without losing information. In this paper, two computerized testing methods for classification into categories are discussed: Computerized Classification Test (CCT) under Item response theory (IRT) and CCT under Stochastic Curtailment (SC). In a post-hoc simulation study, using the item scores on the School Attitude Questionnaire Internet (SAQI) of a large sample of Flemish students (N = 4211), the CCT methods were compared with respect to respondent burden, given minimized differences in proportion of correct classifications. Both CCT under IRT and CCT under SC showed the potential to reduce the respondent burden while maintaining 90% concordance with the full-length questionnaire. CCT under SC required fewer items (20% of the total item set) than CCT under IRT (37%), but in the process made larger misclassifications in terms of item ordering.Show less

Leiden University Student Repository

Documents

In Collections

Comparison of Computerized Classification Tests under Item Response Theory and under Stochastic Curtailment for Three-Category Classification. Shortening the SAQI for assessing low, moderate and high-risk behavior