[us-commits] [ehb54/ultrascan3] 246fd4: extrap_c0: allow the outlier QC to discard more th...

[emre brookes]

  Branch: refs/heads/ehb54-issue-920
  Home:   https://github.com/ehb54/ultrascan3
  Commit: 246fd4b747a296ab696167e6e4f479499ad5f412
      https://github.com/ehb54/ultrascan3/commit/246fd4b747a296ab696167e6e4f479499ad5f412
  Author: ehb54 <brookes at uthscsa.edu>
  Date:   2026-07-20 (Mon, 20 Jul 2026)

  Changed paths:
    M us_somo/develop/include/us_hydrodyn_saxs.h
    M us_somo/develop/src/us_hydrodyn_saxs_iqq_extrap_c0.cpp
    M us_somo/develop/src/us_hydrodyn_saxs_iqq_extrap_c0_script.cpp

  Log Message:
  -----------
  extrap_c0: allow the outlier QC to discard more than one curve

Motivated by Mattia's alpha-syn evaluation: he identifies TWO contaminated curves
(conc 2.41176 as an outlier in MW-vs-concentration, and 0.694246 with an
overestimated MW), while the QC discards at most one and nominates a third
(1.81792). With more than one bad curve present, a single pass judges every curve
against a trend that still contains the other offender, so the nomination itself
can be unreliable.

Detection is now repeated on the CLEANED set while curves keep qualifying, bounded
by a new outlier_max limit. Default 1 reproduces the previous behavior exactly
(and is what the GUI uses, since the dialog is unchanged); the headless driver
exposes it as the `outlier_max <int>` control-file directive so the effect can be
evaluated. Each discard is logged with its ordinal, and hitting the cap while a
further curve still qualifies is reported rather than silently ignored. The
regression-plot red "x" marks the first (strongest) discard; the log records all.

Verified: all 9 regression baselines byte-identical with the default, so nothing
changes unless the limit is raised. On the 7-curve alpha-syn set with the limit
raised, pass 2 nominates 0.694246 -- one of the two curves Mattia flagged -- but
it is correctly kept, failing the systematic gate at 64% one-sided (needs >=70%).

Co-Authored-By: Claude Fable 5 <noreply at anthropic.com>



To unsubscribe from these emails, change your notification settings at https://github.com/ehb54/ultrascan3/settings/notifications