Changes in version 0.3.0 (2026-03-30)                  

New Features

  - Survey-weighted model support: Added support for svyglm and svycoxph
    models from the survey package
      - svyglm: Survey-weighted GLM models (logistic, Poisson, Gaussian)
        with ns() or bs() splines
      - svycoxph: Survey-weighted Cox proportional hazards models with
        ns() or bs() splines
      - Uses spline-coefficient extraction approach to avoid
        model.matrix contrasts issues with survey objects
      - Supports both single spline terms and interaction terms (ns(x) *
        group)
      - Survey-adjusted variance-covariance matrix (sandwich estimator)
        is properly used for confidence intervals
  - Moved splines from Suggests to Imports for direct basis construction
    in survey model path

                        Changes in version 0.2.2                        

Bug Fixes

  - Fixed error when plotting Cox/GLM models with categorical covariates
      - extract_spline_data() now properly preserves factor levels when
        creating prediction data
      - Previously caused "contrasts can be applied only to factors
        with 2 or more levels" error
      - Affects models like coxph(Surv(time, status) ~ ns(x, df=4) + sex
        + age_group, data=dat)

                 Changes in version 0.2.1 (2025-11-19)                  

Bug Fixes

  - Fixed lines being drawn outside Y-axis limits when confidence
    intervals exceed ylim range
      - Lines now properly stop when they go out of bounds instead of
        extending beyond axis limits
      - Each line (main curve, lower CI, upper CI) is handled
        independently
      - If lower CI goes below ylim, only that line is clipped while
        upper CI continues if still in range
      - Ensures clean, professional-looking plots that respect specified
        axis boundaries

                 Changes in version 0.2.0 (2025-10-24)                  

Major Improvements

  - Y-axis scaling: Implemented proper ratio scale display for HR/OR/RR
    when log_scale=FALSE
      - Shows actual ratio values (e.g., 0.5, 1, 2) instead of log
        values
      - Automatically selects appropriate breaks based on data range
      - Removes trailing zeros from labels (1.50 → 1.5)
  - X-axis improvements:
      - Fixed X-axis tick visibility and direction (now properly
        pointing downward)
      - Adjusted plot limits to ensure ticks are always visible
  - Histogram alignment: Base of histogram now correctly aligns with
    secondary Y-axis 0%
  - Internal refactoring: Simplified data handling by always using log
    scale internally for ratio metrics
      - log_scale parameter now only affects Y-axis label display
      - More consistent and predictable behavior

Bug Fixes

  - Fixed missing Y-axis labels in GAM interaction plots
  - Corrected reference line position (always at y=0 for log scale,
    which represents ratio=1)
  - Fixed ylabel column missing in extract_spline_interaction() output

                 Changes in version 0.1.1 (2025-09-30)                  

Bug Fixes

  - Fixed Y-axis tick marks display issue in interaction plots
  - Fixed tick marks protruding from axes when histogram is shown
  - Corrected secondary Y-axis scale for "Percent of Population" in
    interaction plots
  - Fixed X-axis positioning with floating axis for histogram display
  - Improved axis tick alignment for both single and interaction plots

Documentation

  - Updated GAM survival examples to use recommended weights parameter
    format
  - Removed unnecessary logo reference from README
  - Clarified that GAM Cox models should use time ~ s(predictor),
    weights = status format

                        Changes in version 0.1.0                        

Initial Release

Major Features

  - Unified interface for visualizing spline effects from GAM and GLM
    models
  - Support for multiple model types:
      - GAM models from mgcv package with s(), te(), ti() smooth terms
      - GLM/LM models with ns() and bs() splines from splines package
      - Cox proportional hazards models from survival package
  - Automatic detection of:
      - Model type and family
      - Spline terms
      - Interaction variables
  - Support for various outcome types:
      - Hazard Ratios (HR) for Cox models
      - Odds Ratios (OR) for logistic models
      - Rate Ratios (RR) for Poisson models
      - Effects for linear/Gaussian models

Visualization Features

  - Publication-ready ggplot2 output
  - Customizable confidence intervals:
      - Dotted lines (default)
      - Ribbon/shaded style
  - Built-in histogram showing data distribution
  - Reference point marking with automatic SE = 0
  - Support for interaction terms with by-variable
  - Log scale option for ratio outcomes
  - Customizable axis labels and limits

Technical Features

  - Automatic handling of different spline basis functions
  - Proper reference value centering with SE = 0
  - Support for both Surv() and weights methods in GAM Cox models
  - Limited support for pspline() in Cox models

Known Limitations

  - pspline() in Cox models has limited support due to internal
    structure
  - Recommend using ns() or bs() with Cox models for optimal results