Changes in version 0.4.0                        

New Features

  - n_routes parameter in si_estim() - Users can now specify the number
    of transmission routes to model (integer >= 2, default 4).
    Previously this was fixed. The number of mixture components fitted
    is 2*n_routes - 1 for the normal distribution and n_routes for the
    gamma distribution.

  - wind parameter in si_estim() - Adds a window censure interval
    parameter (default 1) to si_estim(), giving users control over
    interval censoring in the likelihood calculation.

  - compare_n_routes() function - New model-selection utility that fits
    si_estim() for each value of n_routes from 2 up to n_routes_max and
    reports AIC and BIC for each fit. Helps identify the optimal number
    of transmission routes from the data.

Documentation

  - Added quick start guide vignette (quick_start_guide.Rmd) providing
    an introductory walkthrough of si_estim() and compare_n_routes().

  - Added window parameter vignette explaining the wind censoring
    parameter and its effect on serial interval estimation.

                        Changes in version 0.3.1                        

New Features

  - Convenience plotting function - Added plot_si_fit_result() which
    accepts si_estim() output directly, handling weight aggregation
    automatically for both normal and gamma distributions

Bug Fixes

  - Fixed logical operator in weighted_var() (| changed to || for scalar
    comparison)
  - Fixed inverted denominator formula in weighted_var() documentation
  - Added defensive guards in weighted_var() for degenerate inputs
    (fewer than 2 values, zero weights)
  - Fixed na.rm handling in weighted_var() to also filter NA weights
  - Removed unreachable dead code in calculate_si_probability_matrix()
  - Fixed ggplot2 deprecation warning (size changed to linewidth in
    plot_si_fit())
  - Fixed ggplot_build() type check in tests for newer ggplot2 versions
  - Added match.arg() validation for dist parameter in
    integrate_components_wrapper()

API Changes

  - Reduced exported functions from 18 to 5. Internal helper functions
    (f0, flower, fupper, f_norm, f_gam, conv_tri_dist,
    integrate_component, integrate_components_wrapper, wt_loglik,
    weighted_var, calculate_si_probability_matrix,
    create_day_diff_matrix, calculate_truncation_correction) are now
    marked as internal. They remain accessible via mitey::: if needed.

Documentation

  - Fixed incomplete @param r documentation in conv_tri_dist()
  - Added explicit Author and Maintainer fields to DESCRIPTION for
    R 4.5.2 compatibility
  - Updated .Rbuildignore to exclude .DS_Store, .Rhistory, ..Rcheck, and
    .claude directories

                        Changes in version 0.3.0                        

New Features

  - Multiple restarts for EM algorithm - Added n_starts parameter to
    si_estim() to run the EM algorithm from multiple starting points and
    select the best result based on log-likelihood. This helps avoid
    local optima when initial values are far from the true parameters
    (fixes #7)

  - Convergence diagnostics - Added tol parameter to si_estim() for
    early stopping when parameters stabilize. Returns converged and
    iterations in output

  - Log-likelihood output - si_estim() now returns loglik for model
    comparison

  - CITATION file - Added proper citation information with Zenodo DOI
    and methodology paper reference

Improvements

  - Expanded test coverage with new tests for
    generate_synthetic_epidemic(), plot_si_fit(), convergence
    diagnostics, and multiple restarts
  - Updated R version requirement to R >= 4.0.0
  - Fixed dependency table in README (ggplot2 is required, brms is
    suggested)
  - Added codecov integration for test coverage reporting

                 Changes in version 0.2.0 (2025-09-02)                  

  - CRAN release - mitey is now available on CRAN!
  - Updated function names for consistency (rt_estim() →
    wallinga_lipsitch())
  - Enhanced documentation and examples
  - Improved package metadata and CRAN compliance
  - All features from 0.1.0 maintained with refinements

                        Changes in version 0.1.0                        

  - Initial release
  - Implements Vink et al. (2014) method for serial interval estimation
    (si_estim())
  - Implements Wallinga & Lipsitch (2007) method for time-varying
    reproduction number estimation (rt_estim(), rt_estim_w_boot())
  - Includes comprehensive validation against historical datasets
  - Supports both Normal and Gamma serial interval distributions
  - Provides bootstrap confidence intervals and visualization functions
  - Developed to support epidemiological analysis in Ainslie et al.
    (2025)
  - Complete documentation with four vignettes demonstrating usage and
    validation