Spline Plot for GAM and GLM Models

Create ggplot2 visualizations of smooth or spline effects from GAM and GLM models. Supports Linear, Logistic, Poisson, and Cox models with interaction terms. Handles GAM smooth terms (s(), te(), ti()), GLM splines (ns(), bs()), and Cox pspline().

Usage

splineplot(
  fit,
  data,
  xvar = NULL,
  by_var = NULL,
  refx = NULL,
  term_index = 1,
  bins = 12,
  xlim = NULL,
  ylim = NULL,
  show_hist = NULL,
  log_scale = FALSE,
  ci_level = 0.95,
  show_ref_point = TRUE,
  colors = NULL,
  ribbon_ci = FALSE,
  xlab = NULL,
  ylab = NULL,
  ylab_right = "Percent of Population"
)

Arguments

fit

A fitted model object (gam, glm, lm, coxph)

data

The data frame used to fit the model

xvar

Character string specifying the variable name for x-axis (default: first spline term)

by_var

Character string specifying the interaction variable (default: auto-detect from model)

refx

Reference value for the x variable (default: median)

term_index

For GAM with multiple smooth terms, which term to plot (default: 1)

bins

Number of bins for histogram (default: 12)

xlim

X-axis limits (default: range of x variable)

ylim

Y-axis limits (default: auto-determined, e.g., c(0.25, 2.0) for HR/OR/RR)

show_hist

Logical, whether to show histogram (default: TRUE)

log_scale

Logical, whether to use log scale for OR/RR/HR (default: FALSE)

ci_level

Confidence interval level (default: 0.95)

show_ref_point

Logical, whether to show reference point marker (default: TRUE)

colors

Named vector of colors for by_var levels

ribbon_ci

Logical, whether to use ribbon style for CI (default: FALSE, uses dotted lines)

xlab

Custom x-axis label (default: xvar name)

ylab

Custom y-axis label (default: auto-determined based on model type)

ylab_right

Custom right y-axis label for histogram (default: "Percent of Population")

Value

A ggplot2 object

Examples

# Create sample data
set.seed(123)
n <- 200
x <- rnorm(n, mean = 50, sd = 10)
lp <- -0.05*(x - 50) + 0.001*(x - 50)^2
y <- rbinom(n, 1, plogis(lp))
dat <- data.frame(x = x, y = y)

# GLM with natural splines
library(splines)
fit_glm <- glm(y ~ ns(x, df = 4), family = binomial(), data = dat)
p <- splineplot(fit_glm, dat)
#> Using 'x' as x variable
#> Using refx = 49.41 (median of x)

# \donttest{
# GAM example (requires mgcv)
if (requireNamespace("mgcv", quietly = TRUE)) {
  fit_gam <- mgcv::gam(y ~ s(x), family = binomial(), data = dat)
  p2 <- splineplot(fit_gam, dat)
}
#> Using 'x' as x variable
#> Using refx = 49.41 (median of x)

# Cox model example (requires survival)
if (requireNamespace("survival", quietly = TRUE)) {
  time <- rexp(n, rate = exp(lp/2))
  status <- rbinom(n, 1, 0.8)
  dat$time <- time
  dat$status <- status
  fit_cox <- survival::coxph(survival::Surv(time, status) ~ ns(x, df = 4),
                              data = dat)
  p3 <- splineplot(fit_cox, dat)
}
#> Using 'x' as x variable
#> Using refx = 49.41 (median of x)
# }