Updates to plotting scripts for CMS analysis by rbarrue · Pull Request #116 · HephyAnalysisSW/GOLLUM

rbarrue · 2026-04-18T12:25:58Z

Fixing a bunch of issues with plotting scripts + implementing new functionality.

Commit messages are self-explanatory.

delphes

…LUM into devel-sbi-rs-hepgpu2

Copilot

Pull request overview

This PR updates several CMS analysis plotting utilities by enabling additional systematic variations, extending configurable process/feature handling for prefit plots, and enhancing postfit/PDF plotting outputs (including new data-fit handling in the postfit binned plotting script).

Changes:

Enable detailed JES component variations (including era-split regrouped sources) in the variations registry.
Improve CLI/config behavior for prefit plotting (process list defaults, clearer --features help, Asimov handling fix).
Adjust postfit/PDF plotting scripts (layout/saving behavior, axis ranges, output formats, and adding an attempt to build data histograms from config for data fits).

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 11 comments.

Show a summary per file

File	Description
`user/robert/variations.py`	Enables many JES component variations (but currently leaves grouping inconsistent).
`user/robert/prefit_sys_binned.py`	Extends default processes, fixes `doAsimov` typo usage, improves CLI help/override behavior.
`plot/postfitsys/compare_fits.py`	Updates plot limits and restructures script entrypoint; changes layout/save calls.
`plot/postfitsys/compare_correlations.py`	Changes layout/save calls for comparison/difference heatmaps.
`plot/postfitsys/binned_fromfit.py`	Adds data-fit handling via sample factory + data histogramming; adjusts ratio plot labeling/range logic.
`plot/pdf/pdf_eigenmode_plots.py`	Tweaks labels/styling and adds SVG outputs for eigenmode plots.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-18T12:30:00Z

+   'jes_EC2'      : {'sys_up': 'CMS_scale_j_Regrouped_EC2_up',         'sys_down': 'CMS_scale_j_Regrouped_EC2_down'},
+   'jes_HF'       : {'sys_up': 'CMS_scale_j_Regrouped_HF_up',          'sys_down': 'CMS_scale_j_Regrouped_HF_down'},
+   'jes_relbal'   : {'sys_up': 'CMS_scale_j_Regrouped_RelativeBal_up', 'sys_down': 'CMS_scale_j_Regrouped_RelativeBal_down'},
+


The JES total variations (jes_total_EtaBin0..5) were removed from variations, but syst_groups['JESTOTAL'] still lists them. Any code that iterates syst_groups and expects these variation keys to exist will now fail (e.g. KeyError / missing-variation behavior). Either re-add the jes_total variations or update syst_groups to reference the new JES components you enabled here.

Suggested change

# Compatibility aliases kept so legacy syst_groups entries such as

# syst_groups['JESTOTAL'] can still resolve these variation keys.

'jes_total_EtaBin0': {},

'jes_total_EtaBin1': {},

'jes_total_EtaBin2': {},

'jes_total_EtaBin3': {},

'jes_total_EtaBin4': {},

'jes_total_EtaBin5': {},

Copilot · 2026-04-18T12:30:01Z

-    'jes_total_EtaBin5'        : {'sys_up': 'CMS_scale_j_Total_EtaBin5_up',      'sys_down': 'CMS_scale_j_Total_EtaBin5_down'},
+   'jes_b'        : {'sys_up': 'CMS_scale_j_FlavorPureBottom_up',      'sys_down': 'CMS_scale_j_FlavorPureBottom_down'},
+   'jes_c'        : {'sys_up': 'CMS_scale_j_FlavorPureCharm_up',       'sys_down': 'CMS_scale_j_FlavorPureCharm_down'},
+   'jeq_g'        : {'sys_up': 'CMS_scale_j_FlavorPureGluon_up',       'sys_down': 'CMS_scale_j_FlavorPureGluon_down'},


The variation key jeq_g looks inconsistent with the surrounding JES flavor keys (jes_b, jes_c, jes_q) and likely is a typo. If downstream tooling expects a jes_g (or similar) naming pattern, this will prevent selecting/grouping the gluon component correctly; consider renaming the key to match the established jes_* prefix.

Suggested change

'jeq_g' : {'sys_up': 'CMS_scale_j_FlavorPureGluon_up', 'sys_down': 'CMS_scale_j_FlavorPureGluon_down'},

'jes_g' : {'sys_up': 'CMS_scale_j_FlavorPureGluon_up', 'sys_down': 'CMS_scale_j_FlavorPureGluon_down'},

Copilot · 2026-04-18T12:30:01Z

+is_data_fit = ("data" in args.fit_result)
+
+# grabbing sample factory to create data histograms downstream
+# may need to move this further down to use pieces of code
+# that get things that I need
+if is_data_fit:
+    from common.yaml_loader import _resolve_features_list


is_data_fit = ("data" in args.fit_result) infers fit type from the fit-result filename/path. This is brittle (e.g. directories containing “data”, or a data fit file not containing that substring). Prefer an explicit CLI flag (e.g. --is-data-fit) or infer from the fit JSON/config (e.g. presence of region['data'] / a fit arg stored in the JSON).

Copilot · 2026-04-18T12:30:01Z

+            )
+
+        for ib, value in enumerate(data_unrolled, start=1):
+            print(f"data histogram: {ib=}, {value=}; {h_total.GetBinContent(ib)=}")


This per-bin print inside the data histogram fill loop will produce very large logs for typical bin counts and regions, which is problematic for batch jobs. Please remove it or guard it behind a debug/verbose option.

Suggested change

print(f"data histogram: {ib=}, {value=}; {h_total.GetBinContent(ib)=}")

Copilot · 2026-04-18T12:30:02Z

        nom = h_total.GetBinContent(ib)
        y   = h_data.GetBinContent(ib)
-        ey  = h_data.GetBinError(ib)   # stat-only = sqrt(y)
+        ey  = h_data.GetBinError(ib)   # stat-only = sqrt(y) \  


The inline comment here ends with a stray backslash and extra spaces. This looks accidental and can be confusing when scanning the code; please remove the backslash/trailing whitespace.

Suggested change

ey = h_data.GetBinError(ib) # stat-only = sqrt(y) \

ey = h_data.GetBinError(ib) # stat-only = sqrt(y)

Copilot · 2026-04-18T12:30:02Z

+    plt.tight_layout()

    os.makedirs(output_dir, exist_ok=True)
-    fig.savefig(os.path.join(output_dir, "correlation_comparison.png"), dpi=150, bbox_inches="tight")
-    fig.savefig(os.path.join(output_dir, "correlation_comparison.pdf"), bbox_inches="tight")
+    plt.savefig(os.path.join(output_dir, "correlation_comparison.png"), dpi=150, bbox_inches="tight")
+    plt.savefig(os.path.join(output_dir, "correlation_comparison.pdf"), bbox_inches="tight")
    plt.close(fig)


Same as above: using plt.tight_layout() / plt.savefig() here depends on the global current figure. Prefer fig.tight_layout() and fig.savefig(...) to avoid saving/layouting the wrong figure if this code is reused in a context with multiple active figures.

Copilot · 2026-04-18T12:30:02Z

+print(f"{default_binning=}")
+print(f"{var_name=}, {var_edges_default=}")


The debug print(f"{default_binning=}") / print(f"{var_name=}, ...") will add noisy output for every run. Consider removing these or gating them behind a verbosity/debug flag (or a logger) to keep script output stable for batch usage.

Copilot · 2026-04-18T12:30:03Z

            h_ratio_down.SetBinContent(ib, 1.0)

+    if is_data_fit:
+        max_dev = max_dev_data


When is_data_fit is true, max_dev is overwritten with max_dev_data, discarding the uncertainty-band driven max_dev computed above. This can clip the MC uncertainty band (and the plot range) if the postfit uncertainty is larger than the data statistical error. Use max_dev = max(max_dev, max_dev_data) instead of replacing it.

Suggested change

max_dev = max_dev_data

max_dev = max(max_dev, max_dev_data)

Copilot · 2026-04-18T12:30:03Z

+    plt.tight_layout()

    os.makedirs(output_dir, exist_ok=True)
-    fig.savefig(os.path.join(output_dir, "correlation_difference.png"), dpi=150, bbox_inches="tight")
-    fig.savefig(os.path.join(output_dir, "correlation_difference.pdf"), bbox_inches="tight")
+    plt.savefig(os.path.join(output_dir, "correlation_difference.png"), dpi=150, bbox_inches="tight")
+    plt.savefig(os.path.join(output_dir, "correlation_difference.pdf"), bbox_inches="tight")
    plt.close(fig)


Same global-state issue here: use fig.tight_layout() and fig.savefig(...) rather than the plt.* variants to make sure the intended figure is laid out/saved, especially if these utilities are imported and used programmatically.

Copilot · 2026-04-18T12:30:03Z

+        ax_top.semilogx(x_vals, pdf_variant, color=color, linestyle=linestyle, linewidth=1.5, label=f'EV({mode_idx})')

        mask = pdf_central > 1e-15
        ratio = np.ones_like(pdf_central)
        ratio[mask] = pdf_variant[mask] / pdf_central[mask]
-        ax_bot.semilogx(x_vals, ratio, color=color, linestyle=linestyle, linewidth=1.5, label=f'EV({mode_idx+1})')
+        ax_bot.semilogx(x_vals, ratio, color=color, linestyle=linestyle, linewidth=1.5, label=f'EV({mode_idx})')
        ax_bot.set_ylim(-1.0,4.0)


The legend labels now use EV({mode_idx}), which will display EV(0) for the first mode (since mode_idx is 0-based). If the plots are meant for presentation/publication, consider labeling as EV({mode_idx+1}) (and keep CLI indices 0-based if desired) or otherwise clarify the indexing convention to avoid confusion.

…alysisSW/GOLLUM into dev-rbarrue_plot_updates

rbarrue · 2026-04-27T10:00:15Z

Closing, as it is superseded by #120, which is equal to this but with a fixed commit history.

schoef and others added 10 commits March 29, 2026 12:21

claude-rs-devel

a497590

Merge pull request #110 from HephyAnalysisSW/devel-sbi-rs

4fdb64a

delphes

delphes

714872e

resolved

5c7fc3a

Merge branch 'devel-sbi-rs-hepgpu2' of github.com:HephyAnalysisSW/GOL…

b820690

…LUM into devel-sbi-rs-hepgpu2

Merge branch 'devel-sbi-rs-hepgpu2' of github.com:HephyAnalysisSW/GOL…

efd76a7

…LUM into devel-sbi-rs-hepgpu2

Fit comparison plots: fixing saving call to use syncer

bcc5482

PDF eigenmode plotting: simpler version for CMS

f696352

Prefit plotting: fixes

1791cd5

Postfit plot script: plot data

98370ff

rbarrue requested review from Copilot and schoef April 18, 2026 12:25

Copilot started reviewing on behalf of rbarrue April 18, 2026 12:26 View session

Copilot AI reviewed Apr 18, 2026

View reviewed changes

rbarrue and others added 10 commits April 20, 2026 11:31

BIT plots (from sbi-pdf)

5577bb9

Merge branch 'devel-sbi-rs-hepgpu2' into devel-sbi-rs

b6e7a83

Postfit plots from binned config + unbinned fit result

42d9f26

Fit comparisons for subsets of parameters

2d55722

Pretty PNN closure plots: changing output directory

3cbd9d5

BIT plots: truth-only plots

50142ba

PDF uncertainty comparison script

71ea02e

Merge plotting-only changes from 'devel-sbi-rs' of github.com:HephyAn…

8670b46

…alysisSW/GOLLUM into dev-rbarrue_plot_updates

BIT option: CLI argument for clipping (off by default)

889e257

Archiving superseded postfit plot script

72b7729

rbarrue force-pushed the dev-rbarrue_plot_updates branch 2 times, most recently from da291d1 to 72b7729 Compare April 23, 2026 09:42

rbarrue added 3 commits April 23, 2026 11:51

Unf***ing commit history

9c420b4

Prefit plots from inputs: deleting broken + moving from Robert's folder

d8addbe

Postfit plot from unbinned config

4bfc0e2

rbarrue mentioned this pull request Apr 27, 2026

Updates to plotting scripts for CMS analysis (fixed commit history) #120

Open

rbarrue closed this Apr 27, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Updates to plotting scripts for CMS analysis#116

Updates to plotting scripts for CMS analysis#116
rbarrue wants to merge 23 commits into
sbi-pdf-CMSfrom
dev-rbarrue_plot_updates

rbarrue commented Apr 18, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

rbarrue commented Apr 27, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

+   # Compatibility aliases kept so legacy syst_groups entries such as
+   # syst_groups['JESTOTAL'] can still resolve these variation keys.
+   'jes_total_EtaBin0': {},
+   'jes_total_EtaBin1': {},
+   'jes_total_EtaBin2': {},
+   'jes_total_EtaBin3': {},
+   'jes_total_EtaBin4': {},
+   'jes_total_EtaBin5': {},

	'jeq_g' : {'sys_up': 'CMS_scale_j_FlavorPureGluon_up', 'sys_down': 'CMS_scale_j_FlavorPureGluon_down'},
	'jes_g' : {'sys_up': 'CMS_scale_j_FlavorPureGluon_up', 'sys_down': 'CMS_scale_j_FlavorPureGluon_down'},

	ey = h_data.GetBinError(ib) # stat-only = sqrt(y) \
	ey = h_data.GetBinError(ib) # stat-only = sqrt(y)

		print(f"{default_binning=}")
		print(f"{var_name=}, {var_edges_default=}")

Conversation

rbarrue commented Apr 18, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

rbarrue commented Apr 27, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants