Skip to content

Latest commit

 

History

History
53 lines (39 loc) · 1.85 KB

CHANGELOG.md

File metadata and controls

53 lines (39 loc) · 1.85 KB
Raw

#Changelog

0.2.1

Changed

  • All nbinom functions are now scipy native, resulting in massive speed improvements for gps()

Fixed

  • pnbinom() provided unstable distribution estimates as previously implemented. New implementation resolves this.

0.2.0

Added

  • New LASSO option to use an NB-GLM backend with L1 regularization
  • Binary conversion forces expansion of summed counts to be compatible with LASSO
  • Binary conversion requires use_counts flag to skip expansion

Changed

  • Added new sections to the readme for all new features
  • setup.py file much cleaner and easy to read/use
  • Improved documentation for GPS and LASSO
  • Final signal detection for LASSO now based on LASSO threshold parameter

0.1.4

Added

  • New LASSO implementation for more DA options
  • Exposed the scipy minimization function to user arguments
  • Default boundaries for the GPS algorithm
  • Data conversion tools for binary matrices and multi-item matrices

Fixed

  • Wrong GPS hyperparameters were being used
  • Errors with BCPNN indexing during signal output

0.1.3

  • Added better error handling in the LBE function to avoid crashing when no objects meet threshold criteria
  • Changed PRR argument to accept fdr_threshold argument which is now passed to lbe()
  • Fixed a bug where returned signal indices were strings instead of ints

0.1.2

  • Added the ability to run a disjoint longitudinal model that does not use cumulative reports during disproportionality analysis
  • Exposed a test_dispersion function from vigipy.utils that determines dispersion and alpha values for any data

0.1.1

  • Fixed bug related to bad logic flow when using signal as a ranking statistic
  • BREAKING: Changed output column names in the signal results to better match function argument conventions

0.1.0

  • Initial Release

Following Semantic Versioning