Using simulation approaches when conducting randomization tests for comparing two groups in the context of experimental studies has been promoted as a beneficial approach for supporting student learning of statistical inference. Many researchers have suggested that the data production process in simulations for the randomization test intuitively connects to the random assignment used in the original study design, thus supporting students' understanding of the logic of inference. Yet, there is little empirical research on how students initially think about the concepts and processes underlying the randomization test as they engage in constructing and using probability models to solve a problem. This work makes a contribution by deepening our understanding of students' reasoning about randomization tests by focusing on a group of three students as they create and use a TinkerPlots model to simulate data and use this data to make a statistical inference. This work adopts a narrative lens through which to view these students' reasoning and modeling activity. We compare and contrast the narratives we constructed for these students along with a narrative we constructed for a statistician. We discuss possible implications for teaching randomization tests for comparing two groups using a modeling and simulation approach.