**The most surprising finding is that weekend extra hours have essentially no relationship with anything else, while weekday daily usage is tied to almost every outcome.** **Daily usage moves with everything (n = 4,500)** | Daily usage vs. | Correlation (r) | |---|---| | Mental health | -0.85 | | Sleep hours | -0.72 | | GPA | -0.71 | | Sleep quality | -0.68 | | Perceived stress | +0.75 | **Weekend extra hours move with nothing** | Weekend extra hours vs. | Correlation (r) | |---|---| | Daily usage | -0.010 | | Mental health | 0.004 | | GPA | 0.012 | | Sleep hours | 0.017 | | Stress | 0.001 | Age and GPA are also unrelated (r = 0.028). **Why this is counterintuitive** - In real observational data, people who use social media heavily during the week usually also use it more on weekends. Here weekend extra hours are statistically independent of daily hours (r ≈ -0.01). - Weekend extra hours also show no link to sleep, stress, mental health or GPA. Daily usage correlates with all of these at |r| of roughly 0.7 to 0.85. - A correlation of -0.85 between usage and mental health is also unusually clean for self-reported student data. Together with the zero correlations, this suggests the variables may be generated or structured around daily usage rather than reflecting messy real behaviour. The dataset card flags it as observational, so treat it with some caution. **Supporting context: the `Overall_Impact` label tracks usage level** | Overall_Impact | Students | Avg daily hours | Avg mental health | Avg GPA | Avg sleep hours | Avg stress | |---|---|---|---|---|---|---| | Beneficial | 3,681 | 4.4 | 84.3 | 3.55 | 7.0 | 11.2 | | Neutral | 654 | 8.5 | 63.4 | 3.00 | 5.5 | 22.3 | | Negative | 165 | 12.3 | 47.1 | 2.54 | 4.2 | 31.1 | Daily usage rises steadily across these three groups, and every outcome moves with it. 82% of students fall in the "Beneficial" group. That label may be defined by usage level, not independently observed. **Implication:** for the wellbeing-outreach questions in this data, limits on weekday daily use are the lever the data points to. Weekend extra hours carry no signal here. This is correlation only and does not show causation.