splineplot 0.3.0

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

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)

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

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

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

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

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

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

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