18  Publication figures

Input Saved model status and coefficient tables.

Action Build and save one figure set per fitted moderator.

Output PNG and PDF files in Rdata/figures/.

Next Run the reproducibility checks.

source(here::here("Scripts", "00_packages.R"))
source(here::here("Scripts", "01_paths.R"))
source(here::here("Scripts", "06_model_registry.R"))
source(here::here("Scripts", "11_plotting_orchard_like.R"))
source(here::here("Scripts", "12_save_figures.R"))

This chapter reads saved model results and generates figures. It does not fit models.

status_file <- here::here("Rdata", "tables", "model_grid_status.csv")
loc_file    <- here::here("Rdata", "tables", "location_effects_all_moderators.csv")
scl_file    <- here::here("Rdata", "tables", "scale_effects_all_moderators.csv")

model_status <- if (file.exists(status_file)) readr::read_csv(status_file, show_col_types = FALSE) else NULL
loc_effects  <- if (file.exists(loc_file))    readr::read_csv(loc_file,    show_col_types = FALSE) else NULL
scl_effects  <- if (file.exists(scl_file))    readr::read_csv(scl_file,    show_col_types = FALSE) else NULL

fitted_ids <- if (!is.null(model_status)) {
  model_status$model_id[model_status$status == "fitted"]
} else character(0)

cat("Models with figures to generate:", length(fitted_ids), "\n")

Models with figures to generate: 7 

18.1 Action: generate figures per moderator

for (i in seq_len(nrow(moderator_grid))) {
  row       <- moderator_grid[i, ]
  mod_id    <- row$model_id
  moderator <- row$moderator
  label     <- row$label
  mod_type  <- row$type

  if (!mod_id %in% fitted_ids) {
    message("Skipping figure for ", mod_id, " (not fitted).")
    next
  }

  loc_dat <- if (!is.null(loc_effects))
    dplyr::filter(loc_effects, model_id == mod_id) else NULL
  scl_dat <- if (!is.null(scl_effects))
    dplyr::filter(scl_effects, model_id == mod_id) else NULL

  if (is.null(loc_dat) || nrow(loc_dat) == 0) {
    message("No location effects for ", mod_id)
    next
  }

  # Build plots based on moderator type
  if (mod_type == "categorical") {
    p_loc <- plot_categorical_location(
      loc_dat |> dplyr::filter(submodel == "location"),
      label
    )
    p_scl <- if (!is.null(scl_dat) && nrow(scl_dat) > 0) {
      plot_categorical_scale(
        scl_dat |> dplyr::filter(submodel == "scale"),
        label
      )
    } else NULL
  } else {
    # For continuous predictors, build a simple coefficient plot
    # (full prediction ribbons require the fitted model object)
    p_loc <- loc_dat |>
      dplyr::filter(submodel == "location", !grepl("^Intercept", term)) |>
      ggplot2::ggplot(ggplot2::aes(x = estimate, y = term)) +
      ggplot2::geom_vline(xintercept = 0, linetype = "dashed", colour = "grey60") +
      ggplot2::geom_pointrange(
        ggplot2::aes(xmin = q2_5, xmax = q97_5), size = 0.6
      ) +
      ggplot2::labs(
        x = "Slope estimate (lnM per unit moderator)",
        y = NULL,
        title = paste("Location slope:", label)
      ) +
      ggplot2::theme_classic(base_size = 12)

    p_scl <- if (!is.null(scl_dat) && nrow(scl_dat) > 0) {
      scl_dat |>
        dplyr::filter(submodel == "scale", !grepl("^Intercept", term)) |>
        ggplot2::ggplot(ggplot2::aes(x = estimate, y = term)) +
        ggplot2::geom_vline(xintercept = 0, linetype = "dashed", colour = "grey60") +
        ggplot2::geom_pointrange(
          ggplot2::aes(xmin = q2_5, xmax = q97_5),
          size = 0.6, colour = "#e06c75"
        ) +
        ggplot2::labs(
          x = "Slope estimate (log sigma per unit moderator)",
          y = NULL,
          title = paste("Scale slope:", label),
          caption = "Positive = greater residual heterogeneity."
        ) +
        ggplot2::theme_classic(base_size = 12)
    } else NULL
  }

  # Save location figure
  fname_loc <- paste0(mod_id, "_", moderator, "_location")
  save_plot_dual(p_loc, fname_loc)

  # Save scale figure
  if (!is.null(p_scl)) {
    fname_scl <- paste0(mod_id, "_", moderator, "_scale")
    save_plot_dual(p_scl, fname_scl, height = 100)

    # Combined figure
    p_combined <- patchwork::wrap_plots(p_loc, p_scl, ncol = 1)
    fname_comb <- paste0(mod_id, "_", moderator, "_location_scale_combined")
    save_plot_dual(p_combined, fname_comb, height = 220)
  }

  message("Saved figures for ", mod_id, " (", moderator, ")")
}

cat("Figure generation complete.\n")

Figure generation complete.

18.2 Output: figure inventory

png_files <- list.files(here::here("Rdata", "figures", "png"), pattern = "\\.png$")
pdf_files <- list.files(here::here("Figures", "pdf"), pattern = "\\.pdf$")

cat("PNG figures:", length(png_files), "\n")

PNG figures: 18 

cat("PDF figures:", length(pdf_files), "\n")

PDF figures: 18 

tibble::tibble(file = png_files) |>
  dplyr::mutate(moderator = stringr::str_extract(file, "(?<=ls_)[^_]+"))