Changes in version 0.2.2 (2026-05-01)                  

Bug fixes

  - Correct default value for arg shift in ecoXCorr() function (from -1
    to 1)

Improvements

  - Correct R code output of the shiny app.

                 Changes in version 0.2.1 (2026-04-01)                  

  - Add a warning when providing ecoXCorr() function with multiple time
    series (i.e. when duplicated dates are found in meteo_data)

                        Changes in version 0.2.0                        

New features

  - aggregate_lagged_intervals() now supports multiple independent time
    series via a new id_col argument. When specified, lagged
    aggregations are performed separately within each group, preventing
    unintended mixing of values across distinct time series (e.g. sites,
    stations, individuals).

Improvements

  - Improved robustness of fit_models_by_lag() when computing R² using
    performance::r2(). Warnings are now handled more gracefully, and R²
    is re-evaluated when possible instead of returning NA.
  - More informative warning messages are collected and returned to the
    user when issues occur during R² computation.

Bug fixes

  - Fixed aggregation interval: The function
    aggregate_lagged_intervals() has been corrected to properly include
    the last day of each interval in the aggregation. Previously,
    intervals were defined as start <= date < end, which systematically
    excluded the final day. This has now been changed to start <= date
    <= end, ensuring that all dates within the intended interval are
    considered when computing summary statistics. This fix guarantees
    that the aggregated results accurately reflect the intervals
    specified by the user. In addition, default for parameter shift has
    been set to 1, i.e. intervals ends the day preceding ref_date (this
    latter being excluded).

  - Fixed an error occurring in fit_models_by_lag() during result
    aggregation: Error in match.names(clabs, names(xi)) : names do not
    match previous names This error occurred when performance::r2()
    failed (error or warning) for some models, leading to inconsistent
    column structures across elements of the results list. The function
    now ensures consistent output structure across all lag windows by:

  - standardizing R² extraction,

  - returning properly formatted values even when warnings occur,

  - preventing failures in do.call(rbind, results).

Notes

  - This update improves stability when fitting large numbers of models
    or using complex model families (e.g. zero-inflated or truncated
    distributions via glmmTMB).