[us-commits] [ehb54/ultrascan3] 800068: SAXS extrap-to-c0: Guinier QC, Primus error propag...
emre brookes
noreply at github.com
Sun Jul 12 07:18:55 MDT 2026
Branch: refs/heads/ehb54-issue-920
Home: https://github.com/ehb54/ultrascan3
Commit: 80006862276e1b0ce576c4de2cdf4269fadaa81d
https://github.com/ehb54/ultrascan3/commit/80006862276e1b0ce576c4de2cdf4269fadaa81d
Author: ehb54 <brookes at uthscsa.edu>
Date: 2026-07-12 (Sun, 12 Jul 2026)
Changed paths:
M us_somo/develop/include/us_hydrodyn_saxs_iqq_extrap_c0_regplot.h
M us_somo/develop/src/us_hydrodyn_saxs_iqq_extrap_c0.cpp
M us_somo/develop/src/us_hydrodyn_saxs_iqq_extrap_c0_regplot.cpp
Log Message:
-----------
SAXS extrap-to-c0: Guinier QC, Primus error propagation, regplot GCV note
Three follow-ups to the GCV penalized-slope extrapolation:
- Guinier quality check: after producing the curve, fit ln I vs q^2 over
the low-q region (refined so q_max*Rg <= 1.3) and report Rg, I(0) and the
fit R^2 as a ground-truth-free QC signal (flagged dark-red if R^2 < 0.9).
This was the decisive real-data test showing the GCV fit beats per-q OLS
and almerge at low q; now surfaced to the user on every extrapolation.
- Primus error propagation: below the merge point (the extrapolated
region) report the weighted-fit intercept standard error instead of the
reference curve's sd, so the low-q extrapolation carries its own (with
GCV, conservative/upper-bound) uncertainty; at/above the merge the output
is the reference and keeps the reference sd.
- Regression viewer: annotate that automatic GCV slope regularization was
in effect (effective slope dof), so it's clear the fit line reflects the
globally-smoothed slope rather than the raw per-q intercept.
Verified (exact re-implementation of the new Guinier helper): on the real
GCV curve it gives Rg=39.2 A, R^2=0.998 over 59 points (q*Rg=1.25).
Review-only; not compiled here.
Co-Authored-By: Claude Opus 4.8 <noreply at anthropic.com>
To unsubscribe from these emails, change your notification settings at https://github.com/ehb54/ultrascan3/settings/notifications
More information about the us-commits
mailing list