Changes in version 3.0.2 (2026-03-02)                  

  - Computation of HCCMEs is now faster by using a crossproduct approach
    rather than constructing a full omega matrix.
  - Changed the default HCCME from HC3 to HC1 (i.e., hc = 1).
  - Fixed an issue where standard errors were returning NA for certain
    edge cases.

                 Changes in version 3.0.1 (2026-02-16)                  

  - Added the lwd and legend parameters to the S3 plot() method for
    UnDiDObj which controls line width and legend locations,
    respectively. Removed the default titles of plots.
  - Also made other minor aesthetic changes to the default plots, such
    as adding horizontal bars to the ends of the event study plot
    confidence bands (like the letter I).
  - Removed "none" as an aggregation option for staggered adoption.

                 Changes in version 3.0.0 (2026-01-26)                  

  - Added new aggregation methods for staggered adoption: "sgt", "time",
    and "none".
    
      - "sgt" computes subaggregate ATTs for each treatment-time and
        post-treatment period pair, further separated by silo.
      - "time" computes subaggregate ATTs grouped by time since
        treatment.
      - "none" skips subaggregate ATT computation and directly computes
        the aggregate ATT from the underlying differences.

  - undid_stage_three() now returns an object of class UnDiDObj,
    providing a unified S3 interface.
    
      - Added S3 methods: print(), summary(), coef(), and plot().

  - Added support for event study plots via the plot() method for
    UnDiDObj objects.

  - Added randomization inference p-values for subaggregate ATTs.

  - Fixed a miscalculation of jackknife standard errors and p-values.

  - Expanded weighting options for ATT computation: "none", "diff",
    "att", and "both".
    
      - "diff" applies weights when computing subaggregate ATTs.
      - "att" applies weights when aggregating subaggregate ATTs.
      - "both" applies weighting at both stages.
      - "none" disables weighting entirely.

                 Changes in version 2.0.0 (2025-05-31)                  

  - start_time and end_time columns are now stored as strings in the
    format specified by date_format column in the empty_diff_df CSV
    file.
  - Fixed as.Date() calls in undid_stage_two(), undid_stage_three(), and
    plot_parllel_trends() to ensure compatability for R versions 4.0.0
    and onwards.

                 Changes in version 1.0.2 (2025-04-27)                  

  - Made compatible for R versions 4.0.0 onwards.

                 Changes in version 1.0.1 (2025-02-24)                  

  - Fixed the miscalculation of standard weights during stage two under
    common adoption scenarios.

                 Changes in version 1.0.0 (2025-01-23)                  

  - Initial CRAN submission.
  - Updated installation instructions.

                        Changes in version 0.1.0                        

  - Initial development version (not yet submitted to CRAN).
  - Provides a framework for implementing difference-in-differences with
    unpoolable data.