v0.4
v0.4.0 (12 March 2017)
This release contains a number of new features and bug fixes.
New Features
- GARD documentation is now hosted on Read-the-docs (#3, #13, #19)
- Continuous integration testing on Travis-CI (#4)
- Add timezone option to handle different timezones in the training/prediction data for the GEFS option (#18)
- Default transform was changed to "No Transform" (#21)
- New option for using training data mean/std to normalize prediction data. This is particularly useful when the prediction data record is quite short. (#22)
- Improved minimum value handling in normalization (#25)
- New option for passing the prediction straight through GARD without any statistical modification (#27)
- Added new
interactive
option to allow users to specify what command line behavior they want (#31) - Improvements to output files, variable names are now sensible (#33)
- New
agg_method
option (GEFS Only) to allow the method (sum or mean) that GARD will aggregate over the time indices in the prediction data (#35) - Weighted averaging (GEFS Only) #37
- Allow for users to cache/reuse coefficients from the pure regression option (#39)
- Experimental post-processing utilities inside of GARD (#49)
- Experimental QQ-Normal transform (#51)