Plot matched facet effects or posterior predictions across MFRMs

Usage

# S3 method for class 'mfrm_extended_comparison'
plot(
  x,
  style = c("paired", "difference"),
  facet = NULL,
  draw = TRUE,
  palette = c("accessible", "mono"),
  title = NULL,
  caption = NULL,
  show_title = TRUE,
  show_notes = TRUE,
  show_labels = metric %in% c("effects", "infit", "outfit"),
  text_scale = 1,
  point_size = 2.5,
  metric = c("effects", "expected_score", "variance", "infit", "outfit", "probability",
    "person"),
  category = NULL,
  ...
)

Arguments

x: A descriptive extended-model result from compare_mfrm().
style: "paired" plots comparison against reference with an equality line. "difference" plots comparison minus reference against their mean.
facet: Panels to display; NULL includes all. For effects and group indices these are facet names; for row moments the panel is "Selected ratings"; for probabilities it is "Score k" for category k.
draw: FALSE returns plot data without opening a device.
palette: "accessible" uses blue points and a dashed grey reference; "mono" uses dark grey points. Meaning does not depend on colour.
title, caption: NULL uses the default text; a string replaces it; "" removes it. Interpretation notes remain in the saved plot data.
show_title, show_notes: Show the title and caption, respectively. FALSE hides text without deleting its metadata.
show_labels: Show matched IDs beside the points. Defaults to TRUE for effects and group indices, FALSE for rating-level moments and probabilities. Consult the retained table when labels are hidden.
text_scale: Positive multiplier for text size.
point_size: Positive point size in ggplot millimetres; base graphics use the corresponding relative size (2.5 is the default).
metric: "effects" compares centered facet locations (default). "expected_score", "variance", "infit", "outfit" and "probability" require saved response_diagnostics in compare_mfrm(). These compare the same posterior predictive target in original score units, squared score units, descriptive mean squares or probabilities. "person" requires saved person_scores and compares EAPs centered at each fitted population mean, in logits. No difference interval is drawn.
category: Numeric score categories to display with metric = "probability"; NULL includes all categories.
...: Unused.

Value

Invisibly, an mfrm_plot_data object with source effects, checks, interpretation notes and a text alternative. as_ggplot() supports further styling without estimation. Both renderers respect the display controls.

Details

Points are centered at the unweighted mean within each complete matched facet. These displays compare fitted summaries, including shrinkage, rather than testing differences or choosing a better model. The equality/zero line is a descriptive reference. No difference intervals or limits of agreement are computed. Unavailable rows remain in plot data. Select facet or use show_labels = FALSE for crowded displays.

Predictive quantities are not centered. Each row label maps through data$response_events and the comparison's responses$rows, which retain both original input row numbers. Infit/Outfit use selected rows in each group; posterior conditioning retains all observed source ratings. Smaller residual indices do not establish better prediction or model adequacy. There are no ordinary fit cutoffs, tests or automatic ranking.