Skip to contents

Plot Correlated Posterior Densities

Source: R/methods_plotting.R
plot_correlated_densities.Rd

Creates a 2D density plot for two correlated parameters from a Stan model fit. Axis limits are computed independently for X and Y using quantiles for a tight data-driven window, with optional padding. Ellipses mark specified probability levels under a bivariate normal approximation.

Usage

plot_correlated_densities(
  x,
  n_grid = 100,
  level_ellipses = c(0.5, 0.8, 0.95),
  quantile_range = c(0.001, 0.999),
  padding = 0.05,
  correlation_method = c("pearson", "spearman", "kendall")
)

Arguments

x

A list-like object that contains a stanfit at x$stan_fit. The fit must have parameters beta_cure_arm and beta_surv_arm.

n_grid

Integer. Number of grid points per axis for the 2D kernel density estimation passed to MASS::kde2d. Default is 100.

level_ellipses

Numeric vector of probability levels for confidence ellipses (bivariate normal contours). Default c(0.5, 0.8, 0.95).

quantile_range

Numeric vector of length 2 with lower and upper tail probabilities used to set axis limits independently. Defaults to the central 99.8%: c(0.001, 0.999).

padding

Numeric scalar (>= 0). Fractional padding to expand each axis range beyond the selected quantiles. Default 0.05 (5%).

correlation_method

Character. The method to use: 'pearson' (default), 'spearman', or 'kendall'. Passed to stats::cor.

Value

A ggplot object representing the joint posterior density.

Details

This helper expects a list-like object x that contains a fitted Stan object under x$stan_fit. The Stan fit must expose posterior draws for two parameters named beta_cure_arm and beta_surv_arm. These are interpreted as:

  • beta_cure_arm: log(OR) for the cure component.

  • beta_surv_arm: log(TR) for the survival component among the uncured.

The function:

  1. extracts posterior samples with rstan::extract(),

  2. computes Pearson, Spearman, or Kendall correlation,

  3. sets independent axis limits via stats::quantile(),

  4. estimates a 2D KDE via MASS::kde2d(),

  5. builds a ggplot2 heatmap with contours and confidence ellipses.

Examples

if (FALSE) {
# Suppose 'fit' is a stanfit with parameters beta_cure_arm and beta_surv_arm:
obj <- list(stan_fit = fit)
p <- plot_correlated_densities(obj,
                               n_grid = 150,
                               correlation_method = "spearman",
                               quantile_range = c(0.005, 0.995))
print(p)
}