Friedland Chapter 9 follow-up: loc/iloc setter, model persistence, Exhibits I and III

[priyam0k]

Summary of Changes

Follow-up to #1109 addressing Henry's review comments on Chapter 9 - Bornhuetter-Ferguson Technique. Three items:

• loc/iloc setter: replaced the tri * 0 + arr trick in the a priori helper with d.iloc[0, 0] = np.asarray(vec, dtype=float).reshape(d.shape).
• model persistence: the XYZ development pattern is now built as a Pipeline, saved with to_pickle, and recalled with cl.read_pickle instead of refitting, so Chapter 9 reuses the Chapter 7 estimators.
• Exhibits I and III: added Exhibit I (U.S. Industry Auto) and Exhibit III (U.S. PP Auto, four changing-conditions scenarios) next to the existing Exhibit II (XYZ). Notebook runs I to II to III with reconciliation asserts and outputs embedded.

Exhibit I reconciles exactly (IBNR 25,609,761 / 27,086,611 reported / paid). Exhibit III reconciles reported and paid for the Steady-State and Increasing Claim Ratios scenarios, and paid for all four.

Related GitHub Issue(s)

refers #588, follows #1109

Additional Context for Reviewers

• Because the text rounds CDFs to 3 decimals, I feed the rounded CDFs into BornhuetterFerguson via DevelopmentConstant(style="cdf"). Without it the unrounded CDF drifts about 18k at U.S. Industry Auto magnitudes.

• Rounding conventions differ per exhibit, matching the Friedland spreadsheets: Exhibit I rounds the LDF then cumulates (full-precision percentages), Exhibit III cumulates raw factors then rounds the percentages to 3 decimals.

• Known data issue: the two case-outstanding scenarios (friedland_uspp_auto_increasing_case, friedland_uspp_increasing_claim_case) do not reconcile on the reported basis because those samples carry uncorrected data (salexanian's corrected CSVs on Reconcile data against Friedland's text used in Exam 5 #588 were not merged). The notebook flags this, and asserts cover reported for the two clean scenarios plus paid for all four. A separate PR to merge the corrected CSVs would let us tighten those asserts.

• Chapter 9 is a notebook, so asserts run at build (execute_notebooks: force) and a failure surfaces as a warning, per the Chapter 11 convention.

• I passed tests locally for code (pytest) and documentation (jb build docs --builder=custom --custom-builder=doctest)

---

Note

Low Risk
Documentation-only notebook changes with executable asserts; no production library or API surface changes in this diff.

Overview
Expands Friedland Chapter 9 (chapter_9.ipynb) from primarily Exhibit II (XYZ) to all three textbook exhibits, with reconciliation asserts and refreshed outputs.

Exhibit I (U.S. Industry Auto) is new: Chapter 7 development (3-year simple average, rounded LDFs/CDFs), Chapter 8 a priori expected claims, BornhuetterFerguson driven via DevelopmentConstant(style="cdf"), plus unpaid/IBNR tables and asserts against printed Exhibit I.

Exhibit II keeps the XYZ workflow but applies review follow-ups: a priori vectors are set with d.iloc[0, 0] = ... instead of zeroing via multiplication, and the Chapter 7 XYZ development selection is built as a Pipeline, saved with to_pickle, and reloaded with cl.read_pickle before transform (instead of refitting inline).

Exhibit III (U.S. PP Auto) adds four changing-conditions scenarios via a shared pp_bf_scenario helper (70% expected claim ratio, 5-year simple development, rounded % unreported/unpaid folded into effective CDFs). The notebook documents known sample-data gaps for two case-outstanding scenarios and asserts reported IBNR totals only where data is clean, plus paid IBNR for all four.

^{Reviewed by Cursor Bugbot for commit 95e5176. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Pyright Type Completeness

View the full pyright --verifytypes output for this commit

Project (full chainladder package, at this PR's head): 14.5% of exported symbols fully typed (185 / 1272)

	Known	Ambiguous	Unknown	Total
Project (head)	185	106	981	1272

Other symbols referenced but not exported by chainladder: 13

	Known	Ambiguous	Unknown	Total
Other (head)	3	1	9	13

Symbols without documentation:

• Functions without docstring: 313
• Functions without default param: 0
• Classes without docstring: 10

Patch (exported symbols added or changed by this PR): no exported symbol type-completeness changes detected.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.50%. Comparing base (d81951d) to head (95e5176).

Additional details and impacted files

@@              Coverage Diff              @@
##           experimental    #1112   +/-   ##
=============================================
  Coverage         90.50%   90.50%           
=============================================
  Files                91       91           
  Lines              5289     5289           
  Branches            671      671           
=============================================
  Hits               4787     4787           
  Misses              359      359           
  Partials            143      143           

Flag	Coverage Δ
unittests	90.50% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 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.