Plot exploratory groups and imputation sensitivity

Inspect within-sample separation, individual feature profiles, or pairwise co-membership across imputations using existing clustering results.

Usage

# S3 method for class 'mfrm_clusters'
plot(
  x,
  type = c("silhouette", "profile"),
  feature = NULL,
  labels = NULL,
  draw = TRUE,
  preset = "standard",
  ...
)

# S3 method for class 'mfrm_imputed_clusters'
plot(x, ids = NULL, labels = NULL, draw = TRUE, preset = "standard", ...)

Arguments

x: An object returned by mfrm_cluster() or mfrm_cluster_imputed(), as appropriate.
type: For a single partition, "silhouette" or "profile".
feature: A single selected feature name, required for type = "profile". Numeric features show means and medians in their original units; categorical features show within-group proportions in the original level order (including unused factor levels).
labels: Whether to display entity IDs on silhouettes or heatmaps. The default displays them for at most 50 entities. No entities are sampled when labels are hidden. Profile plots always label groups and levels.
draw: Draw the plot when TRUE; FALSE only returns plotted values.
preset: Plot style: "standard", "publication", "compact", or "monochrome".
...: Reserved for future use; additional arguments are rejected.
ids: For imputation heatmaps, an optional character vector of distinct entity IDs in the desired display order. Selection affects only the view, not clustering or the denominator. By default all IDs appear in input order.

Value

Invisibly, a ggplot object. plot_data() extracts the plotted table or matrix, title, subtitle, legend, and excluded IDs. Heatmaps also retain the displayed IDs and the number of imputations. Profile plots also retain observations: original-unit entity values and group/exclusion status, suitable for custom distribution plots; see plot_data(). as_ggplot() converts imputation co-membership heatmaps using the saved matrix, ID order, label choice, colours and imputation count. The default and component = "matrix" retain the complete view and its fixed zero-to-one scale. Unavailable cells have both grey fill and crosses, distinguishing them from zero even in monochrome. Metadata, including excluded IDs, remain available with plot_data(). No values are recomputed or renormalized when IDs are selected. Use ggplot2::labs(title = NULL, subtitle = NULL, caption = NULL) to hide annotations while retaining the source data. Silhouette conversion retains negative widths, saved order and the overall mean reference, with group labels independent of colour. Numeric profiles retain original-unit means/medians and counts; circles and triangles are offset vertically to show coincident values without adding intervals. Categorical profiles retain the original category order, unused levels, group counts and a fixed zero-to-one proportion scale. Default and component = "table" preserve the complete selected view; categorical profiles also accept component = "matrix". These conversions do not recluster, select groups, or estimate uncertainty.

Details

No model or clustering is refitted. Silhouette widths describe separation in the fitted sample, not stability or probabilities. The dashed line is the overall mean silhouette; excluded entities have no silhouette.

Feature profiles describe one partition; numeric summaries have no confidence intervals. For an imputed result, inspect a completion with plot(x$analyses[[1]], type = "profile", feature = "ExperienceYears"). Cluster numbers must not be averaged across imputations.

Imputation heatmaps show the fraction of all supplied imputations in which each pair belongs to the same group, on a fixed zero-to-one scale. Grey cells are unavailable pairs involving excluded entities, not zero co-membership. These fractions describe sensitivity to imputations under fixed settings, not membership probabilities, sampling stability, or a consensus partition. Rows and columns follow input order or explicit ids; no hierarchical clustering is performed. PAM is nonhierarchical and these plots do not provide a dendrogram. Use mfrm_cluster(..., method = "hierarchical") for a separate hierarchical analysis and its dendrogram.

Session plot defaults

See mfrmr_visual_diagnostics for options(mfrmr.plot_preset), argument precedence, and appearance when saving or reusing a plot.

Examples

# Supplementary workflow: run explicitly to keep routine checks brief.
# example("plot.mfrm_clusters", run.dontrun = TRUE)
if (FALSE) { # \dontrun{
if (requireNamespace("cluster", quietly = TRUE)) {
  raters <- data.frame(Rater = paste0("R", 1:6),
    ExperienceYears = c(1, 2, 3, 12, 13, 14),
    Specialty = factor(rep(c("Language", "Science"), each = 3)))
  groups <- mfrm_cluster(method = "pam", mfrm_features(raters, "Rater",
    c("ExperienceYears", "Specialty")), k = 2)
  plot(groups)
  plot(groups, type = "profile", feature = "ExperienceYears")
  plot(groups, type = "profile", feature = "Specialty")
  values <- plot_data(plot(groups, draw = FALSE))
  values$table
}
} # }