The sample above shows 15 of the 8,500 rows (one per `user_id`). I left out `bedtime_phone_minutes` because it is flagged as person-identifying, and the same goes for the pipeline metadata columns. **What the sample shows** - **Demographics:** ages run from 18 to 56. It includes female, male and one non-binary user. - **Occupation and chronotype:** it covers Corporate 9-to-5, Healthcare / Shift Worker, Freelance / Creative and Remote Tech, and all three chronotypes (Morning Lark, Intermediate, Night Owl). - **Apps:** it spans TikTok/Reels, Instagram/Reddit, YouTube, Messaging/Chat, Streaming and News/Reading. - **Sleep debt mix:** 8 rows are Mild Deficit, 6 Moderate Debt and 1 Optimal Recovery. There is no Severe Sleep Debt row. **How well it represents the full table** The full dataset's category counts come from the category summary I pulled: - Moderate Debt: 4,462 rows (about 52%) - Mild Deficit: 2,004 rows (about 24%) - Optimal Recovery: 1,387 rows (about 16%) - Severe Sleep Debt: 647 rows (about 8%) The sample therefore leans toward Mild Deficit, has too few Optimal Recovery rows and misses Severe entirely. It is a reasonable illustration of the columns and value formats, but not a proportional slice. I can pull a stratified sample, with rows from each category in proportion to its size, if you want one. **Patterns visible in the rows** - Rows with fewer sleep hours, such as 4.45 to 5.54, are the Moderate Debt ones. They tend to have more alarm snoozes (4 to 5) and higher fatigue scores (3.9 to 5.8). - Rows with 7+ hours are mostly Mild Deficit or Optimal Recovery, with 0 to 3 snoozes and fatigue scores around 1 to 3. - This matches the category-level averages. Severe debt averages 3.76 sleep hours, 5.9 snoozes and a fatigue score of 9.6, while Optimal Recovery averages 8.35 hours, 0.8 snoozes and a fatigue score of 1.1. This is observational data, so these patterns show association, not cause.