Using Fixed‑Effects Panel Models to Assess Within‑Person Variation in Smartphone Unlocks From Continuous Sensing

Traditional screen time research is dominated by cross-sectional designs that conflate between-person differences with within-person processes, limiting causal interpretation of associations with wellbeing. Smartphone unlocks provide objective, timestamped markers of digital engagement that are amenable to intensive longitudinal analysis. This preregistered study (OSF: osf.io/qjxy6) evaluates smartphone unlock frequency as a within-person outcome for fixed-effects panel models and examines its associations with academic stress, physical activity, and weekend status using continuous sensing data from the StudentLife study (47 undergraduates; 3,290 person-days). Fixed-effects panel regressions estimated within-person associations between daily unlocks per hour and time-varying predictors, controlling for individual and calendar-day effects. Unlocks showed substantial within-person variability across students (coefficients of variation typically exceeding 25%), satisfying methodological requirements for fixed-effects estimation. The longitudinal structure supports the use of unlock frequency as a feasible within-person outcome in smartphone sensing studies, and time-varying physical activity showed a small, non-significant negative association with unlocks. The preregistered main effect of physical activity was not supported, whereas exploratory analyses suggested weaker stress-unlock associations on more active days. Days coded as academically stressful showed higher within-person unlock rates than non-stress days, whereas weekends showed lower unlock rates.