By Madeline Hunt (University of Illinois at Urbana-Champaign) and Vimal Rao (University of Illinois at Urbana-Champaign) and Kelly Findley and Mia Petrie (University of Illinois at Urbana-Champaign)
Information
Introductory statistics courses should support students’ ability to “interpret what graphs do and do not reveal” (GAISE, 2016). This stems beyond traditional introductory level graphs (e.g., box plots, scatter plots) to those “statistically-based results reported in popular media” (GAISE, 2016). To support assessment and instruction of these goals, we developed a 32-item, forced-choice assessment based on previous work examining students’ thinking via cognitive interviews (Hunt et al., 2025; Petrie et al., 2025). The Critical Literacy for Evaluating Authentic Representations of Data (CLEAR Data) Assessment is designed to measure students’ authentic visual data literacy, i.e., their ability to interpret what graphs do and do not reveal. To evaluate the items we developed, we recruited 1084 students from 8 different universities for a large nationwide test. Student response data was analyzed with a 2-parameter logistic (2PL) unidimensional item response theory (IRT) model with fixed guessing parameters based on the number of response options. Results suggest that several items were extremely difficult for students, a result consistent with findings from Hunt et al. (2025) and Petrie et al. (2025), but that no items had significant misfit. Model fit statistics indicated good overall fit (RMSEA = .046; SRMSR = .068; CFI = .92; TLI = .91), and empirical reliability was 0.92. These results provide validity evidence related to the internal structure of the assessment that (when considered in combination with other types of validity evidence) can serve to support the following claim: students’ scores from the CLEAR Data assessment can be interpreted as a measure of their ability to interpret what graphs do and do not reveal. These results support the use of the CLEAR Data assessment to support statistics and data science instruction aligned with the GAISE 2016 College Report. Students were recruited from 8 different colleges and universities. These institutions were intentionally sampled to provide a diverse sample of institutional types and students. They include large R1 universities as well as small private liberal arts colleges; they cover all geographic regions of the US, and courses varied in size from a small 20-person class to a large 500-person class.