Detecting Demographic Influences on Measures of Spatial Ability with Rasch Tree Analysis

This study applied Rasch modeling to examine the psychometric properties of two spatial assessments, the Mental Rotation Test (MRT) and Object Assembly (OA), and used group comparisons to explore demographic patterns across the MRT, OA, and Surface Development (SD). Participants were 123 students in Grades 3-8 recruited from university-affiliated summer programs. For the MRT and OA, most items showed acceptable model-data fit, with infit mean square values generally within the expected range of 0.70 to 1.30, although a small number of items displayed evidence of underfit or overfit. Wright maps indicated generally appropriate alignment between item difficulty and participant ability, supporting the usefulness of these measures for distinguishing low-to-moderate levels of spatial skill. Group comparisons based on Rasch-derived theta estimates showed a substantial overlap across gender and race groups on the MRT and OA, with small and non-significant differences. SD raw-score comparisons showed similarly modest subgroup differences. To evaluate measurement invariance at the item level, separate Rasch tree analyses were conducted for the MRT and OA using gender and race as potential splitting variables. Methodologically, the study contributes by combining Rasch-based evaluation of item functioning with exploratory subgroup visualization to examine how spatial assessment performance may vary across demographic groups while maintaining consistency with the psychometric assumptions of the model. Overall, the findings support the value of Rasch-based methods for evaluating the measurement quality of spatial assessments and demonstrate that limited mean group differences at the overall score level can coexist with meaningful item-level noninvariance.