Changes in version 2.1.0                        

New features

  - Parallel bootstrap support is migrated from the snowfall backend to
    the future and future.apply frameworks, enabling cross-platform
    parallel execution.
  - Internal parallel helper functions are introduced to provide a
    unified interface for serial and parallel evaluation.

Improvements

  - Bootstrap routines are refactored to use the new internal parallel
    helpers, simplifying the code base and removing backend-specific
    logic.
  - Additional safeguards are implemented for parallel execution,
    including validation of the number of workers and protection against
    exporting excessively large globals.
  - Documentation and examples are updated to reflect the new parallel
    infrastructure.

Bug fixes

  - Minor robustness improvements in numerical routines, including safer
    handling of near-singular matrix inversion using an SVD-based
    pseudoinverse.
  - Small fixes and consistency improvements across internal helper
    functions.

                 Changes in version 2.0.0 (2026-02-27)                  

Breaking changes

  - The internal architecture is fully refactored to an R6 design.
  - Direct access to internal slots is no longer supported.
  - Users relying on previous S3-style behaviour may need to update
    their code.

New features

  - The PSTR class now follows an R6 design with reference semantics.
  - The main modelling steps are implemented as class methods.
  - Exported wrapper functions remain available for backward
    compatibility.
  - The command line output is redesigned using the cli package.
  - The print layout is modernised using knitr::kable.
  - Separate print views are available for summary, tests, estimation,
    and evaluation.
  - Internal getters and setters are added to support a clean R6
    workflow.

Improvements

  - Evaluation and bootstrap routines are rewritten under the R6
    structure.
  - Result objects are stored with consistent naming (tv, ht, wcb_tv,
    wcb_ht).
  - Numerical robustness is improved in near-singular matrix inversion
    cases.
  - An SVD-based pseudoinverse is used when necessary.
  - Bootstrap routines use deep cloning to avoid overwriting the
    original object.
  - Internal print helpers are reorganised into dedicated private
    methods.

                        Changes in version 1.3.0                        

  - fix the errors in R version 4.
  - change documentation.

                        Changes in version 1.2.5                        

  - fix the bug in the function “plot_transition”.
  - fix the bug in the function “plot_coefficients”.

                 Changes in version 1.2.4 (2019-06-03)                  

  - improve the plotting function “plot_transition” and “plot_response”.
  - add new plotting function “plot_coefficients”.
  - change documentation.

                 Changes in version 1.2.3 (2018-12-09)                  

  - fix some bugs found in the previous version.

                        Changes in version 1.2.2                        

  - fix some bugs found in the previous version.

                 Changes in version 1.2.1 (2018-06-01)                  

  - Link to the working paper.

                        Changes in version 1.2.0                        

  - New function “plot_response” which use the ggplot2 and plotly
    packages to plot the curve or surface of the expected response
    (contribution) against some variable and the transition variable.
  - The estimation function “EstPSTR” will return the estimated
    time-invariante individule effects.
  - The width of the separating lines in printing the results can adjust
    itself automatically to fit the width of the console.
  - Rewrite the “version” function.

                 Changes in version 1.1.0 (2017-09-25)                  

  - Better documentation.
  - In the function “EstPSTR”, now the initial value for “gamma” is
    allowed. The user can freely choose either “gamma” or “delta” to
    input for the estimation.
  - Estimation of a linear panel regression model with fixed effects.
  - The initial values can now be chosen automatically and will be
    returned after estimation.
  - New function “plot_transition” which use the ggplot2 package to plot
    the estimated transition function.
  - The package ggplot2 imported.
  - A new data set “sunspot” included.
  - Some bugs fixed.

                 Changes in version 1.0.1 (2017-09-13)                  

  - All the functions
  - Documetation including README, LICENSE and etc.