Skip to content

Beta-Alanine Lab phase 9: audit fixes - #3

Merged
jesse-anderson merged 7 commits into
mainfrom
beta-alanine-phase9
Nov 18, 2026
Merged

jesse-anderson merged 7 commits into
mainfrom
beta-alanine-phase9

Conversation

@jesse-anderson

Copy link
Copy Markdown
Owner

Works through the Beta-Alanine Lab audit of 2026-11-16, one commit per group of findings.

  • 1.2 Days to target and back to normal rank every draw; a percentile among the misses reads as not reached.
  • 1.3 Evidence horizon judged per run of comparable doses; step downs need a study that stepped down.
  • 1.1 Baguet 2009 washout compared over its own 9 weeks (fourth known gap); new 3-week share anchor.
  • 2.1 Harris 2006 ramp fitted week by week (4-week median 14.81 to 14.73, no verdict changes).
  • 2.2 Person factor CV computed so the total spread matches Rezende's 6.6 (0.37, was 0.47).
  • 2.4 to 2.6 Baseline spread caption, sampled half-life range, Stellingwerff step-down ratio anchor (weak fifth gap).
  • 2.3 Ozdemir 2007 and Baguet 2010 archived: Ghent MRS mM is per liter of tissue, x4.3 stands.
  • 3.x Tingling without body mass, generator parity and schedule checks, the 800 mg reason, whole-day rounding, duration flag.

Every new page figure is a claim in claims.csv. Beta-Alanine spec: 77 tests pass locally; conformance, page-load, hub and contrast specs pass.

Days to target and back within normal variation now rank every draw,
with a miss counted as later than any day. A percentile that falls among
the misses reads "not reached while dosing" or "not within the N weeks
shown" instead of an early day. (Audit 1.2)
Each run of comparable doses is covered from its own start by studies
that followed the same kind of run. A step down is shaded unless a study
stepped down the same way, and a leading rest no longer delays the
shading. The tooltip and edge table handle rest days and uncovered runs.
(Audit 1.3)
Baguet's 0.21 mM/wk is linear over the 9-week washout, so the model side
is now its mean loss of 2 mM over 63 days: model-below, the fourth known
gap. A new anchor for the share gone at 3 weeks (31.8%) is inside.
Ozdemir 2007 and Baguet 2010 are recorded as wanted for the MRS unit
basis. (Audit 1.1, 2.3)
…ende's spread

Harris 2006's ramp arm is fitted as 4.0, 4.8, 5.6 and 6.4 g/day, per its
Methods; the 4-week median moves from 14.81 to 14.73 and no verdict
changes. The person factor's CV is computed from the paper-only draws so
the total spread matches Rezende's 6.6 (0.37, was 0.47). The audit trail
prints the computed CV and both 4-week bands. (Audit 2.1, 2.2)
…werff step-down anchor

The total and percent views say the band is the rise only and give the
baseline SD of 8.7. The washout list states the range a lognormal
half-life is drawn from (Yamaguchi 21.8 to 47.6 days). A new
slope-free anchor compares Stellingwerff's two arms at week 8:
model-above, a weak fifth gap. The view uses carnosineGrams instead of a
hardcoded molar mass. (Audit 2.4, 2.5, 2.6, 3.4)
A plain serving with no body mass reads "Needs body mass", and 0.8 g or
more under 10 mg/kg cites the fixed 800 mg result. The generator matches
the engine's unit-rise expression, refuses out-of-order or overlapping
schedules, guards a zero denominator and computes its checks from
unrounded values. A length that is not a whole number of days is named
in the flags, and the duration flag counts the longest unbroken run of
dosing. (Audit 3.1, 3.2, 3.3, 3.5)
Both papers quantify carnosine against an external phantom, so the Ghent
"mM" is per liter of muscle tissue, which the group reads as wet weight.
The x4.3 conversion stands and blancquaert_08 stays model-below. Muscle
density is recorded as an open question, not a figure. (Audit 2.3)
@jesse-anderson
jesse-anderson merged commit 8939d42 into main Nov 18, 2026
2 of 3 checks passed

This branch was previously deployed

1 inactive deployment
github-pages — 8939d42c Deployed Nov 18, 2026 by jesse-anderson via deploy #175
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant