chainladder 0.6.2

patch to 0.6.1

chainladder 0.6.1

Bug fixes:

• Corrected a bug where TailConstant couldn't decay when the contant is set to 1.0
• #71 Fixed issue where ``Pipeline.predictwould not honor thesample_weight` argument

Enhancements:

• #72 Added drop method to Triangle similar to pd.DataFrame.drop for dropping columns
• Added xlcompose yaml templating
• #74 Dropped link ratios now show as ommitted when callinng link_ratio on a Development transformed triangle
• #73 Triangle.grain now has a trailing argument that will aggregate triangle on a trailing basis

chainladder 0.6.0

Enhancements

• Added TailBondy method
• Propagate std_err_ and sigma_ on determinsitic tails in line with Mack for better compatibility with MackChainladder
• Improved consistency between to_frame and __repr__ for 2D triangles.

Bug Fixes

• Fixed a bug where the latest origin period was dropped from Triangle initialization when sure data was present
• resolves #69 where datetime was being mishandled when ingested into Triangle.

chainladder 0.5.5

Bug fixes

• resolves #62 Slicing cdf not working correctly
• resolves #61 groupby not consistent with pandas
• resolves #60 Development.drop_high and drop_low. drop_high and drop_low only drop one observation even if multiple observations match the criteria.
• resolves a bug in TailCurve where the fit_period slicer was not indexing properly
• Better support for the loc property of a Triangle.

Enhancements:

• Resolves #64 support pandas>=1.0.0. chainladder now works with pandas>=1.0.0
• New feature #63 fill values for undefined link ratio. fillna is a new parameter of the Development estimator
• New feature #58 Attachment age support in Tail methods. attachment_age is a new parameter of the TailCurve estimator.
• New feature #59 Add cl.concat functionality
• Added ability to perform arithmetic between a Triangle and an np.ndarray

chainladder 0.5.4

Enhancements

• Resolves #31 Propagate bootstrap process risk. This should make the BootstrapODPSample estimator fully functional.
• Enhanced Benktander and BornhuetterFerguson methods to allow for stochastic apriori picks.

chainladder 0.5.3

Bug fixes

• MackChainladder was incorrectly calculating MSE when dropping link ratios
• Fixed issue where using grain always returned a development triangle even when fed a valuation triangle

Enhancements

• Added __round__ and __pow__ dunder methods
• Warn on triangles with invalid origin/development combinations

chainladder 0.5.2

Bug fix

• patch to column slicers.

chainladder 0.5.1

Enhancements:

• Substantial speed up in triangle initialization for larger triangles
• Fancy loc/iloc indexing now honors the order the user chooses

Bug fixes:

• Mack Std Err for tail fix xonsistent with R v0.2.11 fix
• Valuation triangles should now be JSON serializable
• Bug in expected loss method predict that wasn't rendering prediction results correctly

chainladder 0.5.0

Enhancements:

• Full GPU-support using cupy.

chainladder 0.4.10

v0.4.10

version number