<![CDATA[This study investigates undergraduate students’ performance on a university-level statistics assessment and evaluates the psychometric quality of the instrument using both Classical Test Theory (CTT) and Item Response Theory (IRT). The assessment was administered to 431 students enrolled in engineering and business programs and comprised 16 multiple-choice items selected from an 85-item bank. These items were aligned with four performance indicators related to inferential statistics and regression analysis and were further classified according to cognitive demand and representational format (graphical, tabular, and textual). Descriptive results indicate that the majority of students achieved acceptable levels of performance (scores ≥ 3 on a five-point scale). However, reliability analyses revealed low internal consistency (Cronbach’s α < 0.60), including a negative alpha coefficient for one indicator, suggesting weaknesses in construct validity. IRT analyses further demonstrated that the item bank was disproportionately weighted toward low-difficulty items and that certain constructs—most notably those involving tabular representations—were negatively associated with overall test performance (0.47). In contrast, items requiring part–whole reasoning (0.73) and conceptual understanding (0.70) emerged as the strongest predictors of student success. Collectively, these findings indicate that university statistics assessments should extend beyond procedural computation to foreground conceptual interpretation, proportional reasoning, and meaningful connections across representations. The study underscores the need for improved assessment design that achieves an appropriate balance among item difficulty, discriminative capacity, and cognitive alignment. Future research should replicate these analyses across multiple cohorts and incorporate qualitative approaches to more deeply examine students’ statistical reasoning processes.]]>
