Where the number comes from
92
Of 130 available ratios, the ones whose 95% interval clears 1
Every cell of the grid at the source file's own heavy-user threshold of 3, with a 95% interval from the Katz log method. The bolded rows are the ones whose interval includes 1 — arithmetically correct, and indistinguishable from no difference at all.
| cutoff | ratio | 95% interval | heavy flagged | others flagged |
|---|---|---|---|---|
| ≥ 5 | 1.21× | 1.14 – 1.28 | 1,579 | 936 |
| ≥ 6 | 1.24× | 1.16 – 1.33 | 1,390 | 801 |
| ≥ 7 | 1.27× | 1.18 – 1.37 | 1,227 | 691 |
| ≥ 8 | 1.29× | 1.19 – 1.40 | 1,069 | 592 |
| ≥ 9 | 1.31× | 1.20 – 1.44 | 944 | 514 |
| ≥ 10 | 1.33× | 1.21 – 1.47 | 827 | 444 |
| ≥ 11 | 1.33× | 1.19 – 1.49 | 726 | 390 |
| ≥ 12 | 1.38× | 1.23 – 1.55 | 654 | 339 |
| ≥ 13 | 1.40× | 1.23 – 1.59 | 585 | 299 |
| ≥ 14 | 1.43× | 1.24 – 1.63 | 524 | 263 |
| ≥ 15 | 1.50× | 1.29 – 1.74 | 460 | 220 |
| ≥ 16 | 1.45× | 1.24 – 1.70 | 414 | 204 |
| ≥ 17 | 1.51× | 1.27 – 1.79 | 364 | 173 |
| ≥ 18 | 1.54× | 1.28 – 1.85 | 322 | 150 |
| ≥ 19 | 1.58× | 1.30 – 1.91 | 302 | 137 |
| ≥ 20 | 1.60× | 1.30 – 1.96 | 272 | 122 |
| ≥ 21 | 1.70× | 1.35 – 2.14 | 233 | 98 |
| ≥ 22 | 1.93× | 1.50 – 2.48 | 213 | 79 |
| ≥ 23 | 1.95× | 1.49 – 2.55 | 188 | 69 |
| ≥ 24 | 2.00× | 1.50 – 2.68 | 168 | 60 |
| ≥ 25 | 2.16× | 1.58 – 2.95 | 154 | 51 |
| ≥ 26 | 2.45× | 1.73 – 3.47 | 137 | 40 |
| ≥ 27 | 2.32× | 1.61 – 3.34 | 120 | 37 |
| ≥ 28 | 2.35× | 1.59 – 3.47 | 105 | 32 |
| ≥ 29 | 2.60× | 1.70 – 3.96 | 98 | 27 |
| ≥ 30 | 2.68× | 1.72 – 4.19 | 90 | 24 |
| across the whole grid | |
|---|---|
| cells in the grid | 130 |
| interval clear of 1 | 92 (70.8%) |
| ratio range | 1.15× – 2.97× |
The weak cells cluster at the extremes, where the flagged group shrinks to a few dozen teenagers. That is exactly where the most quotable ratios live.
Screen Time vs Mental Health (ML-ready)Kaggle · kylefengkfeng209/screen-time-vs-mental-health-ml-ready · diambil 2026-07-30
