Introductory data science courses have the potential to provide students from diverse backgrounds skills for working with and reasoning about data. However, what predicts success in these courses remains poorly understood. Here we investigate how students' initial psychological orientation relates to their subsequent engagement and learning. In Study 1, we took an observational approach, analyzing data from 1306 students across 11 institutions using an interactive online textbook. Students' psychological orientation, (e.g., math anxiety, stress expectations) predicted performance on assessments administered throughout the term. In Study 2, we developed and tested an intervention targeting these aspects of students' learning experience among 146 students enrolled in a single course. Preliminary analyses suggest that this intervention shifted students' beliefs about the relationship between stress and learning. This work highlights the promise of combining observational studies with interventions for advancing understanding of how affective and cognitive processes interact to support learning in real-world settings.