At a large university, the overall correlation between hours spent on social media per week and GPA is weak and negative ($r = -0.10$). However, when the data are split by major, both STEM majors and non-STEM majors show positive correlations ($r = +0.25$ in each group) between media time and GPA. Which best explains this apparent contradiction (Simpson's Paradox)?