**The most surprising finding is that weekend extra hours are statistically unrelated to everything else, while every other usage measure is tightly linked to wellbeing.** **What is tightly linked.** Across all 4,500 students, daily social media hours track the outcomes very strongly: | Daily usage vs. | Correlation | |---|---| | Mental health index | -0.85 | | Sleep hours | -0.72 | | GPA | -0.71 | | Sleep quality | -0.68 | | Perceived stress | +0.75 | The chain continues from there. Sleep and mental health correlate at +0.77, stress and mental health at -0.77, and GPA and mental health at +0.66. The `Overall_Impact` label follows the same pattern: | Overall impact | Students | Avg daily hours | Avg mental health | Avg GPA | Avg sleep (hours) | |---|---|---|---|---|---| | Beneficial | 3,681 | 4.4 | 84.3 | 3.55 | 7.0 | | Neutral | 654 | 8.5 | 63.4 | 3.00 | 5.5 | | Negative | 165 | 12.3 | 47.1 | 2.54 | 4.2 | **What is not linked.** `Weekend_Extra_Hours` has essentially zero correlation with the other measures: | Weekend extra hours vs. | Correlation | |---|---| | Daily usage | -0.010 | | Mental health | +0.004 | | GPA | +0.012 | | Sleep hours | +0.017 | You would expect heavy weekday users to also use more on weekends, and weekend binges to cost some sleep or wellbeing. Neither shows up. Age and GPA are also uncorrelated (+0.03). **Why this matters.** - Only weekday-style daily hours seem to carry the effect, so a recommendation to limit weekend use has no support here. - The weekday relationships are unusually strong (r ≈ -0.85) for self-reported student data. The weekend and age columns are flat, while the other measures line up almost as one dose-response curve. That pattern looks more like a generated or rule-driven dataset than organic behaviour, even though the dataset card labels it observational and organic. - Even if the data is real, the correlations are observational and cannot show that usage causes the outcomes. A student who is already stressed or sleeping badly might also use social media more. I did not check whether the weekend result holds within subgroups such as platform or academic level, so I can only say it holds overall.