← Back to the question · Front page
Client name: Bill
No, based on the analysis below, and within its assumptions and limits.
Do you have a similar question? Join the waitlist.
Each of your 15 six-week blocks, split within its epoch into the ones when you slept more than that epoch's middle block, and the rest.
If more sleep helped, the lean dots would climb from left to right and the fat dots would fall. If anything they lean the other way, in both the split above and the test below, but by no more than chance would make.
For p-values, corrections for testing many questions, effect sizes and the smallest detectable differences, see The statistics behind this below.
How much you slept, on average, across each block.
How your body composition changed over each block.
Registered on 2026-10-25, before any sleep number was set beside any scan. Neither measure moved with sleep by more than chance would make. The closer of the two was fat mass: about a 1-in-3 chance of a link that strong by luck alone (p = 0.29, raw); allowing for the six questions tested together on the same history, the corrected value is 1.00.
One scan's own error is about 1.6 lb of lean, which is why the unit is a six-week block, and why the smallest change this test could have seen is as large as it is.
For p-values, corrections for testing many questions, effect sizes and the smallest detectable differences, see The statistics behind this below.
Per outcome: slope of the change per 14 days on mean nightly sleep, both centred within epoch; permutation p over 20,000 shuffles of blocks within epoch; a rough range from 2,000 bootstrap resamples of blocks within epoch; the smallest detectable slope (the 95th percentile of the shuffled slopes); Holm across the six questions tested together on the same history. The superseded run built Epoch C's blocks from that epoch's scans alone, which moved its first block's start; the registration walks every scan in order.
| Run | Outcome | Blocks | Slope, lb / 14 d / h | Rough range | Detectable | p | Holm (6) |
|---|---|---|---|---|---|---|---|
| registered | lean mass | 15 | -0.98 | -2.59 to +0.53 | 2.03 | 0.392 | 1.00 |
| registered | fat mass | 15 | +0.86 | -0.08 to +2.27 | 1.38 | 0.291 | 1.00 |
| illness periods dropped | lean mass | 12 | -1.43 | -4.16 to +0.30 | 2.60 | 0.315 | — |
| illness periods dropped | fat mass | 12 | +1.16 | -0.21 to +2.94 | 1.97 | 0.330 | — |
| every scanner offset removed | lean mass | 15 | -1.07 | -2.47 to +0.16 | 1.85 | 0.277 | — |
| every scanner offset removed | fat mass | 15 | +0.52 | -0.61 to +2.23 | 1.51 | 0.572 | — |
| blocks sharing a scan | lean mass | 19 | +0.32 | -1.06 to +2.20 | 1.90 | 0.759 | — |
| blocks sharing a scan | fat mass | 19 | +0.30 | -0.72 to +1.58 | 1.30 | 0.686 | — |
| first run, superseded | lean mass | 15 | -1.07 | -2.59 to +0.26 | 1.96 | 0.336 | — |
| first run, superseded | fat mass | 15 | +0.74 | -0.29 to +2.15 | 1.44 | 0.392 | — |
Detectable: A real change this big would just pass the test, about half the time. One about 1.4 times as big would pass it about 4 times in 5. A smaller real change could easily go unseen, so 'no change found' means 'no change this big'.
Asked on 2026-10-27, after the results above had been read. Each is one of the extra checks, not a registered test, and lean mass is the only outcome.
No sign. Over 13 pairs of back-to-back blocks, lean gain in the second moved +0.15 lb per two weeks per extra hour of sleep in the first. The smallest change this test could have seen was about 2.1 lb.
No sign. Blocks averaging 7 hours or more gained +0.15 lb more lean per two weeks than the rest. Only Epoch C had any: every Epoch B block averaged under 7 hours. 7 hours is the adult recommendation (Watson et al. 2015, SLEEP).
No sign. With sleep, weekly sets and the two together in one model, on the 12 blocks that have both, the together term is close to zero.
Corrected together, three at a time (Benjamini-Hochberg): 0.91, 0.91, 0.91. Raw p: 0.91, 0.86, 0.66.
Each one runs on your own export. What you asked is written down with today's date before anything is computed, so the question stays separate from the answer.
Everything on this page uses history you had already lived. The same test, the same code, on six-week blocks that start after the registration (2026-10-25) is the real check. At a scan every two to six weeks that takes a year or more.
Nothing but to keep scanning and logging as you do.
Running from 2026-10-25
The full question puts sleep, sets, protein, calories, steps and alcohol in one model. Only 5 of the 15 blocks have food logged (none from 2025-01 to 2025-09), too few for six levers, so it runs on blocks still to come.
Log food every day in MacroFactor.
Waiting for data
If you have a health-related question about your own data, join the waitlist for your own custom analysis.
Act soon: it is free while we stand this service up and refine it. All we ask in return is permission to publish a results page like the ones we have made for our other clients. We never publish your surname or your date of birth. The page shows a first name or a pseudonym, whichever you prefer, and your age in years and your sex.
Made by PerSci from your DEXADEXA (dual-energy X-ray absorptiometry) is a low-dose X-ray scan that measures body fat, lean mass, bone, and total body weight. More at en.wikipedia.org/wiki/Dual-energy_X-ray_absorptiometry scans and your OuraOura makes a ring worn on the finger that measures sleep, heart rate, heart rate variability and skin temperature overnight, and scores each night. More at ouraring.com ring. It describes a pattern in your data; it is not medical advice.