BOT: Fix #942: Add plot_discrimination() for binary forecasts

[nikosbosse]

Summary

• Adds a new plot_discrimination() function that visualises the discrimination ability of binary forecasts by plotting overlapping density curves of predicted probabilities, stratified by observed outcome level
• Includes 7 tests covering basic usage, faceting, plain data.frame input, missing columns, single-model data, separation verification, and edge cases
• Follows existing plot function conventions (ensure_data.table(), check_columns_present(), theme_scoringutils())

Closes #942

Root cause

No binary-specific plot functions existed in the codebase. The forecast_binary class and example_binary dataset provided the data structures but there was no way to visualise discrimination ability.

What changed

• New file R/plot-discrimination.R with plot_discrimination() — accepts a data.frame/data.table with observed (factor) and predicted (numeric) columns, renders overlapping density plots via ggplot2
• New test file tests/testthat/test-plot_discrimination.R with 7 test cases and 2 visual regression snapshots
• NAMESPACE updated to export the new function

Test plan

• All 7 new tests pass
• Full test suite passes (689 tests, 0 failures)
• R CMD check: 0 errors, 0 warnings, 2 notes (pre-existing)

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.00%. Comparing base (d4bbd6e) to head (8dfd005).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1079      +/-   ##
==========================================
+ Coverage   97.98%   98.00%   +0.02%     
==========================================
  Files          37       38       +1     
  Lines        1984     2006      +22     
==========================================
+ Hits         1944     1966      +22     
  Misses         40       40              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[nikosbosse]

CLAUDE: Approved. Clean implementation of plot_discrimination() that follows all existing package conventions (ensure_data.table, check_columns_present, theme_scoringutils). All 7 tests match specifications and pass. Only minor nitpick: unused @importFrom checkmate assert_data_frame import — harmless but unnecessary.

[nikosbosse]

Here is an example that it produced:

devtools::load_all()
  library(ggplot2)
  p <- plot_discrimination(na.omit(example_binary))
  ggsave("/tmp/plot_discrimination.png", p, width = 8, height = 5, dpi = 150)

I like it, though I think we maybe would want a histogram version as well.

[nikosbosse]

[seabbs]

or? it is the output? Do we want to support input that isn't a forecast binary object? Based on package flow I would assume we are mving towards not doing so.

[seabbs]

lintr really works this badly with standard dply?

[seabbs]

I think this looks pretty good. My main concern is do we want to support non forecast objects here or insist that they are first as_ before use in these functions.