<![CDATA[Instruments composed of items worded in the same direction may be vulnerable to response bias. This study evaluated a modified Computational Thinking Scale (CTS) comprising 19 matched pairs of positively and negatively worded items. An instrument-development design adapted from ADDIE was used. The study population comprised science-track senior high school students in the Surakarta Residency. The sample included 393 Grade XI students from three purposively selected schools. Data were collected using a 38-item, five-category self-report questionnaire. Five experts assessed content validity, yielding coefficients of 0.80–0.95 and a mean of 0.89. Responses were analyzed using the Rasch model in Winsteps 5.7.3.0. Cronbach’s alpha and person reliability were 0.66, with person separation of 1.40, whereas item reliability was 1.00 and item separation was 14.67. Response categories showed ordered observed averages and Andrich thresholds, with outfit MNSQ values of 0.91–1.07. All items met the primary outfit MNSQ criterion, ranging from 0.82 to 1.21. Three algorithmic-thinking items showed statistically significant but small gender DIF contrasts of 0.30–0.35 logits. The hypothesis was partially supported because the instrument showed acceptable item-level functioning but limited person-level discrimination. Future studies should examine wording effects, matched-pair equivalence, dimensionality, and local dependence in broader, more diverse samples.]]>
Copyrights © 2026
