Reconcile data against Friedland's text used in Exam 5

[kennethshsu]

Latest Status:

Chapter	Title	Status
Chapter 6	The Development Triangle as a Diagnostic Tool	Done ✅
Chapter 7	Development Technique	@henrydingliu @salexanian
Chapter 8	Expected Claims Technique	@kennethshsu
Chapter 9	Bornhuetter-Ferguson Technique	@priyam0k
Chapter 10	Cape Cod Technique	@priyam0k
Chapter 11	Frequency-Severity Techniques	@henrydingliu
Chapter 12	Case Outstanding Development Technique	Unassigned
Chapter 13	Berquist-Sherman Techniques	Unassigned
Chapter 14	Recoveries: Salvage and Subrogation and Reinsurance	Unassigned
Chapter 15	Evaluation of Techniques	Unassigned
Chapter 16	Estimating Unpaid Allocated Claim Adjustment Expenses	Unassigned
Chapter 17	Estimating Unpaid Unallocated Claim Adjustment Expenses	Unassigned

Data sources and notes

Description

We should reconcile the examples shown in Friedland's text.

Is your feature request at odds with the scope of the package?

No response

Describe the solution you'd like.

Add tests to cover the examples shown in the text.

Describe alternatives you've considered.

No response

Do you have any additional comments?

Supporting docs, provided by one of our friends, Josh T. Thanks Josh!

WM_Appendix_A.xlsx
WM_Appendix_B.xlsx
WM_Appendix_C.xlsx
WM_Appendix_D.xlsx
XYZ_Insurer_BF_Method.xlsx
XYZ_Insurer_BS_Methods.xlsx
XYZ_Insurer_Case_OS_Method.xlsx
XYZ_Insurer_CC_Method.xlsx
XYZ_Insurer_Dev_Method.xlsx
XYZ_Insurer_EC_Method.xlsx
XYZ_Insurer_Freq_Sev_Methods.xlsx

[kennethshsu]

@henrydingliu FYI!

[henrydingliu]

Yo I am completely noobing. cl.load_sample('friedland_us_industry_auto') gives me a no such file error. Other sample data tables work (like raa).

[henrydingliu]

@genedan did you come across this before?

[henrydingliu]

Chapter 7.zip

Here is the first set of exhibits from Chapter 7. We are missing the geometric average.

[genedan]

@henrydingliu

I ran into a bug, but a different one from yours:

test = cl.load_sample('friedland_us_industry_auto')
Traceback (most recent call last):
  File "/home/ubuntu/anaconda3/envs/cl_dev/lib/python3.12/site-packages/IPython/core/interactiveshell.py", line 3549, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-2-94233cefd2f9>", line 3, in <module>
    test = cl.load_sample('friedland_us_industry_auto')
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ubuntu/pr/chainladder-python/chainladder/utils/utility_functions.py", line 161, in load_sample
    return Triangle(
           ^^^^^^^^^
  File "/home/ubuntu/pr/chainladder-python/chainladder/core/triangle.py", line 143, in __init__
    index, columns, origin, development = self._input_validation(
                                          ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ubuntu/pr/chainladder-python/chainladder/core/base.py", line 83, in _input_validation
    if "object" in data[columns].dtypes:
                   ~~~~^^^^^^^^^
  File "/home/ubuntu/anaconda3/envs/cl_dev/lib/python3.12/site-packages/pandas/core/frame.py", line 4108, in __getitem__
    indexer = self.columns._get_indexer_strict(key, "columns")[1]
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ubuntu/anaconda3/envs/cl_dev/lib/python3.12/site-packages/pandas/core/indexes/base.py", line 6200, in _get_indexer_strict
    self._raise_if_missing(keyarr, indexer, axis_name)
  File "/home/ubuntu/anaconda3/envs/cl_dev/lib/python3.12/site-packages/pandas/core/indexes/base.py", line 6249, in _raise_if_missing
    raise KeyError(f"None of [{key}] are in the [{axis_name}]")
KeyError: "None of [Index(['values'], dtype='object')] are in the [columns]"

So it looks like your version can't find the file whereas mine has trouble handling the data set. Are we both on the same version? I'm on 0.8.24.

I guess more work needs to be done to solve #561, in addition to just updating the documentation. cl.load_sample() has a series of if/else statements to assign things like field names from the files to the arguments of cl.Triangle() and right now the Friedland data isn't part of that.

I'll update the issue with a list of what needs to be done, I'll also bring it up tomorrow as part of our discussion.

[henrydingliu]

@genedan I'm also on 0.8.24. Does it matter that I'm installing via pip?

[kennethshsu]

I see what @henrydingliu sees, I actually don't think these are implemented. See this.

@genedan are you on dev?

[henrydingliu]

I see what @henrydingliu sees, I actually don't think these are implemented. See this.

@genedan are you on dev?

Gene did forgot to add the parameterization. But that's not causing this file not found error.

[genedan]

@kennethshsu I'm running it from source, as opposed to having installed it from PyPI.

It turns out the Friedland datasets didn't make it to PyPI, so if you installed the package from there, they'll be missing, but if you cloned the repo from GitHub, you'll see them.

https://pypi.org/project/chainladder/#files

vs.

https://github.com/casact/chainladder-python/tree/v0.8.24/chainladder/utils/data