New plotPopulationCuts() plots PV distributions with a supplied numeric cut vector, without an IDM object. It supports wide/long PVs, weights, population groups, and the existing density and annotation options. Unrounded interval percentages are available through attr(plot, "population_percentages"), including when percentage labels are hidden.
Supplied cuts must be finite and strictly increasing. Interval shares are calculated from weighted observations separately for each PV, then averaged across PVs; values exactly on a cut enter the upper interval. Density smoothing and display rounding do not change these shares.
plotPopulationCutsIDM() now shares its internal drawing code with plotPopulationCuts(). Its public arguments, IDM cut selection, calculations, and plot appearance are unchanged. The new function has help examples and regression tests covering weighted percentages, cut boundaries, wide/long input, population groups, missing PVs, and rendering.
eatPrep 1.0.12
population_fill now also colors grouped populations in both plot functions: one explicit color applies to all groups, or a vector supplies one color per group in legend order. An incorrect number of colors warns and restores the default palette. Omitting the argument preserves existing defaults; named population_colors takes precedence. Fills, density outlines, legend keys, and grouped percentage labels use the same colors.
Population SDs in both plot functions now apply the variance correction n/(n-1) separately within each PV and population, counting available respondents with positive weights. Corrected PV variances are averaged before taking the square root.
Both plotPopulationCutsIDM() and plotCutsIDM() now display M and SD for each population in the legend (or the subtitle for ungrouped input). Weighted means and descriptive population variances are estimated separately per PV, then averaged across PVs; SD is the square root of the average variance. Use show_population_stats = FALSE to hide them and population_stats_digits to set decimal places. In the rater view, PV data are required.
Explanatory captions are now hidden by default in both plot functions. Set show_caption = TRUE to display the explanations of densities/silhouettes, percentages, and M/SD. This does not hide the statistics or percentage labels themselves.
Population percentages now align with the visible cut intervals in both plot views. Crowded labels shift minimally with thin connection lines, while population rows stay aligned. Placement adapts to figure size, zoom, and reversed score axes.
Population preparation preserves distinct large numeric respondent, PV, and population identifiers, including during missing-data checks.
Population density grids adapt to the bandwidth so narrow, separated distributions remain accurately resolved. Smoothing settings requiring excessive grid resolution produce an informative error.
Percentage tables reserve space above each rating panel based on their rendered text height, keeping percentages clear of curves and cut labels when figures are resized.
The package now requires ggplot2 3.5.0 or newer for independent population and cut colors.
Both population views now show per-population percentages for each panel’s cut intervals. Shares are computed from weighted PV observations and averaged across PVs, independently of density smoothing; equality belongs to the upper interval. Use show_percentages = FALSE to hide the tables, and percentage_digits / percentage_size to format them. Incomplete cut panels show an explanatory message. Silhouette height remains a display of distribution shape, not rating stages.
Both population plot views gain population_col to overlay multiple populations and population_colors for named colors. Densities use a common bandwidth and grid, with independent weight normalization and a separate population legend. Background silhouettes share one height factor and continue to show distribution shape only, not rating stages.
plotCutsIDM() gains item_number_size, cut_value_size, and aggregate_label_size to adjust text sizes inside the graph, including plots with residual panels. Defaults preserve the existing sizes.
New plotPopulationCutsIDM() displays a population density estimated from wide- or long-format plausible values, with optional sampling weights and mean, individual, or both sets of IDM cuts.
plotCutsIDM() gains an optional population background through pv_data. The silhouette communicates distribution shape only: its height does not represent rating stages or density-axis values. Enable its explanatory caption with show_caption = TRUE; residual panels stay clear.
Both population views share per-PV density estimation, a common bandwidth and grid, explicit missing-data handling, and adjustable smoothing and appearance. Help examples and the main-functions vignette demonstrate both input layouts and both plot purposes.
eatPrep 1.0.11
improvements
plotCutsIDM() now shows bold numeric cut values next to the vertical cut lines by default in individual rater panels and in the aggregate panel.
Cut value labels are rounded to whole numbers by default; use cut_value_digits to choose a different number of digits after the decimal point.
plotCutsIDM() now shows small bold item position numbers next to the raw grey points in individual rater panels by default.
Cut value and item number labels can be hidden with show_cut_values = FALSE and show_item_numbers = FALSE.
eatPrep 1.0.10
new features
computeCutsIDM() now supports long-format IDM input with explicit rater identifier and rating columns.
computeCutsIDM() gains item_id_col for explicit item identifiers, especially useful when long-format data contain several items with the same difficulty estimate.
computeCutsIDM() supports ordinal rating labels via rating_levels, for example c("1a", "1b", "2", "3", "4").
computeCutsIDM() now reports IDM agreement diagnostics: item-wise modal values, rater-modal correlations, pairwise Cohen kappa summaries, Fleiss kappa, and ICC agreement/consistency.
improvements
computeCutsIDM() now uses missing = "drop" by default so missing ratings do not create smoothed values or contribute to cut computation.
Long-format IDM input is completed to the full item-by-rater grid before smoothing, so omitted item-rater rows are treated like explicit missing ratings.
computeCutsIDM() offers missing = "smooth" for legacy-like smoothing with missing values ignored inside the moving-average window.
computeCutsIDM() now computes cut scores as linearly interpolated boundary crossings of the monotonized curve.
computeCutsIDM() returns additional IDM summary tables in cut_positions_per_person, cut_statistics, and level_statistics.
Cut score labels now reflect the actual boundary. Canonical numeric cuts retain labels such as cut12 and cut23; ordinal and non-canonical cuts use labels such as cut_1a_1b or cut_1a_1b_bound1_3.
plotCutsIDM() can add an aggregate panel with show_aggregate = TRUE; it overlays all monotonized rater curves in rater-specific colors, labels them by rater name by default, and shows the mean cuts from cuts_summary.
plotCutsIDM() now shows the raw rating trajectory and the non-monotonized moving average by default; both can be hidden with show_raw = FALSE and show_smoothed = FALSE. The smoothed line is dashed so that it remains visible when it coincides with the monotonized curve.
plotCutsIDM() can add residual panels with show_residuals = TRUE; residuals are computed as raw rating minus smoothed moving average.
bug fixes
plotCutsIDM() no longer adds aggregate rater lines or rater-name labels to the cut score legend.
compatibility
Existing wide-format calls to computeCutsIDM() remain supported.
The returned object remains compatible with plotCutsIDM().
eatPrep 1.0.9
bug fixes
mergeData() now stops when missing values are found in an ID variable instead of allowing missing IDs to be joined.
mergeData() no longer suppresses diagnostics for non-mbd value conflicts when overwriteMbdSilently = TRUE and another conflict in the same variable involves mbd.
tests
Added regression tests for missing IDs in mergeData() and mixed mbd/non-mbd merge conflicts.
eatPrep 1.0.8
bug fixes
Improved handling of tibble inputs across data preparation workflows.
