Graphs are commonly used in physics learning and professional inquiry. Some argues that being “graphicated†is as important as literate and numerate in contemporary society. The purpose of this study is to examine whether the inclusion of graphs in physics items is associated with high school students’ performance, using the fixed effects model. Our preliminary analysis was performed with a sub-sample (N=200) randomly selected from a large-scale national examination from China, based on their responses on 22 multiple true-false item sets. Each item set was scored at a scale of 0 to 2 (0 when all options were wrong, 1 when some of the options were correct, and 2 when all options were correctly selected). The score on each item set was taken as dependent variable. Two item features were included as independent variables: (1) presence of a graph (ten items with graph and twelve without graph); and (2) level of cognitive demand (content of all the items was analyzed and coded as low, medium, and high by content experts in physics, and we will do ICC to calculate the inter-rater reliability before the formal presentation). In addition, we used a student fixed effects model to control for some covariates related to students (e.g., gender, initial performance level, etc.). Results indicate items that contain a graph were associated with better student performance (β = 0.12, p < 0.01). The interaction term of cognitive demand and graph was insignificant (β = 0.04, p >0.05), indicating the lack of moderation effect in the model. For the formal presentation, we will incorporate the whole sample (N=22, 030). Results from this study help understand whether and how the inclusion of graphs influences students’ performance in high school physics assessments, and provide important implications for research and practice in science assessment.