Skip to content

Latest commit

 

History

History
97 lines (75 loc) · 3 KB

CHANGELOG.md

File metadata and controls

97 lines (75 loc) · 3 KB
Raw

Change Log

All notable changes to this project will be documented in this file per Keep a Changelog. This project adheres to Semantic Versioning.

[Unreleased]

Added

  • When synth errors on a treatment unit, change to redoing the analysis uncaptured so that output/errors are visible to help debugging. Add an option noredo_tr_error to not do this.
  • Option to save all the variable and unit weights by specifying both aggfile_v and `aggfile_w'.

Fixed

  • made trlinediff work with all the graphs.
  • Fixed single_treatment_graphs to work when mod(n units, 10)==0
  • Fixed formatting of help.
  • Renamed calc_RMSPE.ado to calc_rmspe.ado because when net installing on Linux all files are made lower-case!
  • Fixed normalization/scale bug in calc_rsmpe.ado (relative comparisons were correct).

[1.6.0]

Added

  • Help files for the graphing commands

Changed

  • Removed some redundant options in the graphing commands

[1.5.0]

Added

  • The gen_var to easily include the synthetic control estimates in the dataset.

Changed

  • Deprecated keep(file) and replace options (in favor gen_var)

[1.4.0]

Added

  • Added dynamic setting of synth's xperiod and mspeperiod option.

Changed

  • With multiple treatment periods, restrict history the maximim common support. Added option to maintain old behavior (allowing maximal histories in estimation).
  • Relabel the row labels the return matrix to be more intuitive (pre vs post numbers).

[1.3.0]

Added

  • Added dynamic covariates, dynamic donor pool, determinism, parallel support, and more unit tests.

[1.2.0]

Added

  • A separate testing suite aside from usage.do.

Fixed

  • Allowed for gaps in the time variable. The panel should still be strongly balanced with no missing values the units.

Changed

  • Removed "pseudo t-statistic" nomenclature in favor of "standardized effect". This affected some return value names.
  • Update docs.

[1.1.6]

Fixed

  • Recover more gracefully from errors when estimating treatment effects.

[1.1.5]

Fixed

  • Recover gracefully (and report on) optimization errors by donors.
  • Recover (and say why) from a too low max_lead.
  • Update docs.
  • Allow checking version programmatically.

[1.1.4] - 2016-05-02

Fixed

  • Issue with multiple treatment units/per_rmspe_mult
  • Cleanup better after run

Changed

  • Update docs and source code notes.

[1.1.4] - 2016-04-18

Fixed

  • Allow non-consecutive IDs in graphing

[1.1.3] - 2016-04-02

Fixed

  • allow keep files without file extensions
  • make graphs look better on older Stata's

Changed

  • Fixed typos in documentation

[1.1.2] - 2016-03-31

Changed

  • Fixed typos in documentation

[1.1.1] - 2016-03-02

Fixed

  • Allow new additions to work on Stata v12 and v13.

[1.1.0] - 2016-03-01

Added

  • Only use lots of memory when computing CI. By default if # of placebo averages is very large, do a random sample.

Changed

  • Update docs