Skip to contents

Plot Posterior Predictive Checks from Ensemble Predictions

Source: R/plot_model_ppc.R
plot_model_ppc.Rd

Creates professional posterior predictive check (PPC) plots to assess model fit using CSV files from stochastic/ensemble predictions. Generates diagnostic plots across six pages: density overlays, credible interval coverage, calibration scatter plots, Q-Q plots, residuals vs observed, and temporal residual patterns.

Usage

plot_model_ppc(
  predictions_dir = NULL,
  predictions_files = NULL,
  locations = NULL,
  model = NULL,
  output_dir,
  verbose = TRUE
)

Arguments

predictions_dir

Character string specifying directory containing prediction CSV files. Function will auto-discover predictions_ensemble_.csv or predictions_stochastic_.csv files.

predictions_files

Character vector of specific CSV file paths to use. Overrides predictions_dir if provided.

locations

Character vector of specific locations to plot. NULL (default) uses all locations.

model

Legacy: a model object as returned by run_simulation() (any list carrying $results). Not recommended - use CSV-based inputs instead.

output_dir

Directory where PPC plots will be saved. Creates "ppc" subdirectory.

verbose

Logical indicating whether to print progress messages (default: TRUE)

Value

NULL (invisibly). Creates PDF files as side effects.

Details

Creates a 6-page multi-page PDF per output file:

  1. Density overlays of observed vs predicted central (the run's central_method: median for cases and mean for deaths by default) distributions

  2. Interval coverage analysis (50% and 95% intervals vs nominal; the CSV intervals are observation-level predictive intervals for runs made with v0.101.0 or later)

  3. Observed vs predicted calibration scatter plots

  4. Quantile-quantile plots

  5. Residuals vs observed

  6. Temporal residual patterns

Output files:

  • ppc.pdf - Aggregate diagnostics (all locations combined)

  • ppc_ISO.pdf - Per-location diagnostics (e.g., ppc_ETH.pdf, ppc_KEN.pdf)

Examples

if (FALSE) { # \dontrun{
# From predictions directory (auto-discovers CSV files)
plot_model_ppc(
  predictions_dir = "output/3_results/predictions",
  output_dir = "output/3_results/figures"
)

# Specific locations only
plot_model_ppc(
  predictions_dir = "output/plots/predictions",
  output_dir = "output/plots",
  locations = c("ETH", "KEN")
)
} # }