Skip to content

Analysis Page Accuracy and Clarity Improvements - #313

Merged
chvvkumar merged 2 commits into
mainfrom
dev
Sep 21, 2026
Merged

chvvkumar merged 2 commits into
mainfrom
dev

Conversation

@chvvkumar

@chvvkumar chvvkumar commented Sep 21, 2026 •

Copy link
Copy Markdown
Owner

What's new

  • Display plate scale warnings on the Analysis page.

What's fixed

  • Corrected figures and statistics on the Analysis page.
  • Improved accuracy of verdicts and insights.
  • Updated help text for better understanding.

…page

An audit of the Analysis page (#311) found fifteen faults, several of
which put wrong numbers in front of the user. The histogram lost the
frame at the maximum value because the last bin edge was accumulated in
floating point and landed below it; the last bin now ends exactly at the
maximum and counts always sum to the sample size. The Compare verdict
divided by the smaller median while describing the smaller group as
"X% lower", so it printed figures above 100 percent; it now divides by
the larger median. The correlation verdict assumed lower is always
better, which read backwards for detected stars; wording now follows
each metric's polarity and stays neutral for ADU metrics. The Time
Series tooltip named an arbitrary target on multi-target nights and now
lists the targets.

Pixel HFR plotted across rigs with different plate scales now raises a
visible warning on every tab that plots it. The 95% confidence band uses
the Student t value for the sample size instead of a fixed 2.0, so it is
no longer under half its correct width on small samples. Correlation
coefficients report n/a rather than 0.0 when there are too few points or
a constant axis, matching the matrix. Box plot months group by imaging
night like every other grouping, groups dropped for having fewer than
four frames are listed under the chart, and "not enough data" errors
from the API are shown instead of an empty chart.

Queries feeding figures are now ordered, so identical requests return
identical results and the same downsampled points. Outlier fences are
computed once per request instead of re-sorting both axes per point,
skewness uses full-precision inputs, and two unreachable branches are
gone. Help texts for Matrix, Time Series, Granularity and the date range
now describe what the page does. The OpenAPI export and generated types
are refreshed, which also picks up endpoints added since the last export.
Analysis Page Data and Text Corrections
@github-actions github-actions Bot changed the title Dev Analysis Page Accuracy and Clarity Improvements Sep 21, 2026
@chvvkumar
chvvkumar merged commit cc567db into main Sep 21, 2026
12 checks passed
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.

Analysis page: histogram drops the max value, Compare prints over 100% lower, and 13 other findings in analysis.py

1 participant