Skip to contents

Plot Path Coefficients (Caterpillar Plot)

Source: R/plot_coef.R
plot_coef.because.Rd

Creates a 'caterpillar plot' (point and whisker) of all path coefficients from a because model. This visualization provides a precise statistical complement to the plot_dag() overview, allowing for a side-by-side comparison of effect sizes and their Bayesian credibility intervals.

Usage

# S3 method for class 'because'
plot_coef(
  object,
  type = "raw",
  multinomial_probabilities = TRUE,
  color_scheme = "sig_only",
  ...
)

Arguments

object

A because object.

type

Character; either "raw" (default) or "marginal".

  • "marginal": Shows Average Marginal Effects (AME). For categorical predictors, this represents the average shift in the outcome (e.g. probability or counts) associated with a one-category change. This is the recommended scale for comparing cross-model impacts.

  • "raw" (default): Shows the raw structural parameters (betas/rhos) from the JAGS model. Useful for model diagnostics but harder to interpret on the original data scale.

multinomial_probabilities

Logical; if TRUE (default), expands multinomial predictors into a "bundle" of category-specific effects, matching the arcs in the DAG.

color_scheme

Character; color scheme for significance. Options:

  • "sig_only" (default): Discrete Black/Grey scheme. Significant paths (where 95% CI excludes zero) are Black; non-significant are Light Grey.

  • "directional": Switched to a directional Red/Blue/Grey scheme.

  • "monochrome": All effects are rendered in Black regardless of significance.

...

Additional arguments.

Value

A ggplot object. Use standard ggplot2 functions like + ggtitle() or + theme() to further customize the output.

Details

The function automatically sorts coefficients by the Response Variable, effectively grouping all predictors for a given outcome together. This hierarchy makes it intuitive to read 'down' the plot to see what factors contribute most to a specific part of your causal system.

Examples

if (FALSE) { # \dontrun{
# Fit a model
fit <- because(list(Y ~ X + Z, X ~ Z), family = c(Y="binomial", X="gaussian"), data = dat)

# Plot marginal effects for intuitive interpretation
plot_coef(fit, type = "marginal")

# Customizing the plot
library(ggplot2)
plot_coef(fit) + labs(title = "Causal Influence on Wildlife Tolerance")
} # }