Friedland Chapter 6 and half of Chapter 7

[henrydingliu]

Summary of Changes

looking to push the work so far on friedland before the pr becomes toooo large.

adding an index for friedland
adding ch6 and half of 7
adding all to the toc (sitting between user guide and example gallery)

Related GitHub Issue(s)

#588

Additional Context for Reviewers

please visit the experimental version on rtd review content for accuracy and legibility

not checking the box below because i didn't develop the rst locally. but tests are passing on the branch

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

---

Note

Low Risk
Low risk: changes are documentation-only (new friedland section and TOC wiring) plus minor formatting tweaks, with no impact to runtime code paths.

Overview
Introduces a new documentation section docs/friedland/index.rst and adds two new RST chapters (chapter_6.rst, chapter_7.rst) that recreate Friedland exhibits via chainladder doctest examples.

Updates docs/_toc.yml to include the new Friedland section between the user guide and gallery, and makes small formatting-only tweaks to .github/CODEOWNERS and the docs extras list in pyproject.toml.

^{Reviewed by Cursor Bugbot for commit 3901057. Bugbot is set up for automated code reviews on this repo. Configure here.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.19%. Comparing base (3a8c5d5) to head (3901057).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #798   +/-   ##
=======================================
  Coverage   86.19%   86.19%           
=======================================
  Files          86       86           
  Lines        4934     4934           
  Branches      642      642           
=======================================
  Hits         4253     4253           
  Misses        484      484           
  Partials      197      197           

Flag	Coverage Δ
unittests	86.19% <ø> (ø)

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

☔ 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.

[henrydingliu]

will squash and merge so it doesnt flood the entire commit history

[henrydingliu]

open to any and all feedback on the overall flow (some manual coding in the beginning; then more scalable functions towards the end)

also open to any and all wordsmithing on the actual texts in the rst.

[genedan]

Just letting you know I started reading. I'll be reading along between work breaks so it might take a while for full feedback. But what really pops out is the amount of typing you had to do to get the exhibits formed, which you'd mentioned before.

You've revealed that there's still a big gap between getting the numerical results and having them being presentable, which means we really need to work on improving the I/O of the package to get people to use it. I think anyone, including myself would have to do at least as much work as you did to make those summaries.

This is a good candidate for having a summary method that displays the link ratios vertically alongside the column values and ultimate. Something like IBNR might be hard to incorporate due to different companies' field definitions, but I think even a bare bones combo of [origin, link ratio, triangle columns, ultimate] and then having the user have the option to insert their own additional columns/ibnr formula would go a long way.

So I'd vote +1 on an additional summary method and I think we can use these examples as a starting point.

[kennethshsu]

I agree that the amount of "words" is very distracting. I don't want to erase any of the work, and we can probably save the clean up for later, but I'd be perfectly happy with something simple like this.

tri = cl.load_sample("friedland_us_industry_auto")["Reported Claims"] #page 106, part I, triangle data
tri

tri.link_ratio #page 106, part II, age-to-age factors

cl.Development(average="simple", n_periods = 5).fit(tri).ldf_ #page 106, part III, simple average latest 5

I don't think we need to do any intentional rounding to match the text; however, we should perform intermediate rounding to match the text if the rounded numbers are used in a subsequent action.

Honestly, I think this is perfect for a first pass. Let's keep going and see what other methods we are missing so we can get those in. (Still working #272...)

[genedan]

One suggestion for the title. How about "Friedland Walkthrough"? Friedland text sounds like we might be referring to the actual paper, but something along the lines of walkthrough would make it sound more clear that it's a CL application of the text. I'm open to different names too.

[kennethshsu]

"Reproducing Friedland text"?

[kennethshsu]

"Friedland in chainladder-python"?

[henrydingliu]

I agree that the amount of "words" is very distracting. I don't want to erase any of the work, and we can probably save the clean up for later

I'm completely open to other presentation styles. I'm hoping people can take a chapter or a section of a chapter and just crank out something in their style and we can vote on the most beginner friendly.

I don't think we need to do any intentional rounding to match the text; however, we should perform intermediate rounding to match the text if the rounded numbers are used in a subsequent action.

yes, I'm only doing rounding when I need the test output to match the text.

One suggestion for the title. How about "Friedland Walkthrough"?

What about 'Learning Reserving (Friedland)`

Since geomean is so close, I'd like to hold off on merging till that's ready.

[kennethshsu]

"Replicating Friedland's Text"?

[kennethshsu]

Oh oops already have something similar above haha