Column 2: evidence · every driver, item by item
Open data, followed all the way down to one person inside it.
Every piece here starts with a public dataset and ends with a story. The numbers can be checked: any figure set in mono opens its own evidence in a column to the right, so you never have to leave the sentence you are reading to find out whether it is true.
open to work
work
files
The same per-item breakdown, run for all five drivers in the file. If a flat profile were evidence against a mechanism, sleep — eight times stronger, with an obvious mechanism — ought to look different. It does not.
| driver | mean r | sd | spread ÷ size | top-5 share | strongest item |
|---|---|---|---|---|---|
| screen time | +0.0799 | 0.0174 | 0.218 | 30% | 03 |
| sleep quality | +0.2247 | 0.0419 | 0.187 | 28.5% | 19 |
| hours slept | −0.1537 | 0.0330 | 0.215 | 30.4% | 19 |
| weekend midsleep | +0.0878 | 0.0321 | 0.366 | 36.1% | 19 |
| social jetlag | +0.0421 | 0.0245 | 0.581 | 44.7% | 19 |
spread ÷ size is the standard deviation of the 21 correlations divided by their mean, so drivers of very different strength can be compared. Screen time 0.218, sleep quality 0.187. Social jetlag, the weakest driver in the file, is the least uniform of the five — the opposite of what the discarded reading would predict.
the one item that does stand out
Item 19 is the strongest correlate of every sleep measure here, all four of them. That is a real pattern rather than an artefact of one noisy column. The source ships the items unlabelled, so what it asks is not something I can tell you.
| driver | strongest | its r | weakest |
|---|---|---|---|
| screen time | item 03 | +0.1133 | item 06 |
| sleep quality | item 19 | +0.2844 | item 21 |
| hours slept | item 19 | −0.2197 | item 21 |
| weekend midsleep | item 19 | +0.1477 | item 13 |
| social jetlag | item 19 | +0.0866 | item 13 |