v0.6.0

What's Changed

• Allow for zero in the sd for pattern by @AHsu98 in #69
• Joss paper by @saravkin in #36
• Feature/logit sex split by @saalUW in #63
• update tests by @saalUW in #74
• moved splitter to age_splitter testing by @saalUW in #75
• ruff formatting by @saalUW in #76
• Added sex splitting tests by @saalUW in #77
• Refactor some of the core functions by @zhengp0 in #68
• add schema class to abstract prefix functions by @zhengp0 in #71
• Solved same column name issue between data and pattern by @saalUW in #78
• Feature/cat split by @saalUW in #79

Full Changelog: v0.5.1...v0.6.0

v0.5.1

Added

• Count splitting:

  • output_type = 'rate' satisfies the condition that the population weighted mean of the post-split result adds up to the pre-split value.
  • output_type = 'total' satisfies the condition that the SUM of the post-split results adds up to the pre-split value.

• Age splitting columns:

  • Population_total: the sum of age group specific populations across a study.
  • Population_proportion: the normalized population in each post-split group.
  • Implication: can multiply population_proportion by sample study size to get the post-split pseudo sample sizes.

Fixed

• Error messaging in sex splitting if a population is missing.
• Error where pattern and pattern_sd were missing, but actually provided

v0.5.0 (Sex-split)

Changelog

[0.5.0] - 2024-06-17

Added

• Sex splitting API, it requires data, ratio pattern, population. It assumes the ratio is female/male (requires draw mean and standard error).
• Propagating Zeros, there is now an optional argument to the end of the split function in AgeSplitter, propagate_zeros=False which can now be switched to True if there are 0s that you’d like to keep. WARNING This changes standard error of the estimate to be 0 as well. Short reason: we assume a binomial distribution that is strictly positive. Please take this warning seriously as it will crash other modeling processes at IHME (sd == 0).

Changes

• Restructured files, which will change import call in python but should not impact R users.
• Age-splitting API now has “Age” appended to the front of the config files for clarity/ consistency (DataConfig -> AgeDataConfig).
• Error messages. Thanks to the feedback from the users we have changed the error messaging to be more succinct.

Bug-fix

• If a study had an age range fully contained within a single age group pattern it would error with “pattern not found”. This is now rectified and the study should be passed through. Keep in mind that the pre-split and post-split values will be the same.

Full Changelog: v0.4.0...v0.5.0

v0.4.0

What's Changed

• Added functionality for normalizing populations by @AHsu98 in #33
• Bumped version by @AHsu98 in #34
• Fix renormalization for dataframe splitting by @AHsu98 in #35
• Add age split adapted api by @AHsu98 in #37
• Refactor/py pkg update by @saalUW in #39
• Refactor/ihme -J by @saalUW in #40
• Refactor API by @saalUW in #44
• Feature/pattern draws by @zhengp0 in #45
• Fixed example pattern in test_splitter by @AHsu98 in #46

New Contributors

• @saalUW made their first contribution in #39

Full Changelog: v0.3.0...v0.4.0

Refactored, added pattern uncertainty

Refactored the code to be more of a functional style
Added pattern uncertainty propagation
Generally runs faster, draws are not too slow anymore

v0.2.0

What's Changed

• Changed location_id to demographic_id by @AHsu98 in #21
• Updated init.py by @AHsu98 in #18
• Custom fit for simple model by @AHsu98 in #20
• Added rate functionality in high level api by @AHsu98 in #24
• updated to use getter and setter by @AHsu98 in #23
• Reformate demographic_id and refactor disaggregate.py by @AHsu98 in #25

Full Changelog: v0.1.0...v0.2.0

Initial Release

First release to pypi!