Releases: datasciencecampus/synthgauge
v2.2.0
This release adds two new features to the metrics.propensity
module:
- An additional method for estimating pMSE null statistics that only uses the
real data. Eventually, this will become the default behaviour and a warning
is issued if the old permutation-based method is used. Details of the method
can be found in Bowen and Snoke (2021). - The SPECKS metric: a propensity-score-based metric that uses the KS distance.
Details in Bowen et al. (2021).
Many thanks to Claire McKay Bowen and her co-authors!
Full comparison: 2.1.0...2.2.0
v2.1.0
In this minor release, we add a new set of utility metrics from the 2018 NIST
competition. We also fix a couple of small bugs and make some minor improvements
to the documentation.
Full comparison: 2.0.0...2.1.0
v2.0.0
This release hopes to improve the reproducibility of synthgauge
by adding
reliable random number sampling, robust tests, fuller documentation, code
formatting, continuous integration, code styling and bug fixes.
As well as these additions, several parts of the codebase have been refactored
and reorganised to be clearer to users and developers at the expense of
breaking back-compatibility.
A summary of the changes can be found in the change log.
Full comparison: v1.0.0...2.0.0
v1.0.0
This marks the initial release of SynthGauge!
Full Changelog: https://github.com/datasciencecampus/synthgauge/commits/v1.0.0