<![CDATA[Measuring critical thinking skills (CTS) in the topic of atomic structure requires a diagnostic instrument capable of capturing students' reasoning patterns accurately, rather than merely distinguishing between correct and incorrect answers. This study aimed to develop and evaluate a complex multiple-choice diagnostic instrument based on the Rasch model to measure students' CTS on the concept of atomic structure. The instrument comprised ten items developed based on Facione's six dimensions of critical thinking: interpretation, analysis, evaluation, inference, explanation, and self-regulation. Data were collected from 850 senior high school students in Gorontalo, Indonesia, and analyzed using the Partial Credit Model (PCM) approach with WINSTEPS 4.5.5 software. Results indicated that item reliability was very high (0.99; separation = 10.63), while person reliability was moderate (0.61; Cronbach's Alpha = 0.65). Infit and Outfit MNSQ values ranged within the ideal threshold (0.99–1.00) with ZSTD values approaching zero, confirming adequate model fit. Category Probability Curve (CPC) analysis confirmed that response categories functioned sequentially, while the Wright Map demonstrated a progressive difficulty hierarchy from the interpretation to the self-regulation dimension. Differential Item Functioning (DIF) analysis revealed differential difficulty levels based on gender and grade level, particularly for self-regulation items, which proved more challenging for female students and Grade XI students. These findings underscore the importance of considering cognitive factors and learning experience in developing Rasch-based CTS diagnostic instruments for chemistry education]]>
