Skip to contents

Changelog

Source: NEWS.md

deaviz 0.3.0

  • New plot_io_3dfrontier(): the three-dimensional counterpart of plot_io_frontier(). For two chosen inputs and one output it draws the units in three dimensions together with the efficient frontier as a semi-transparent piecewise-linear surface (crs or vrs), colours the units by their efficiency in that three-variable sub-model, and can overlay lambda-weighted peer arrows. The plot is interactive and requires plotly and geometry.

  • Added a test suite validating the two- and three-dimensional frontier geometry against the Benchmarking solver: efficient units must lie on the drawn frontier, no unit may lie above it, and each unit’s radial projection must land on it.

deaviz 0.2.0

  • New dea_panel_data() (with as_dea_panel_data() and a print() method): the panel counterpart of dea_data(). It validates the panel once and stores the column mapping, so plot_panel_io_biplot() and plot_panel_io_parcoo() can be called on the object without repeating inputs/outputs/id/period.

  • New plot_panel_io_parcoo(): traces a chosen input/output or the per-period efficiency of every DMU across time, with the periods as parallel axes; color_by_trend colours each line by the DMU’s overall change (viridis, or a zero-anchored diverging scale with "diverging").

  • New plot_io_frontier(): the classic two-dimensional DEA frontier-and-envelope plot for a chosen input/output pair. Draws the units, the shaded production-possibility set, and the efficient frontier for the selected returns-to-scale (crs/vrs/drs/irs/fdh); colours points by their efficiency in that two-variable sub-model; and optionally overlays the peer (reference) network as lambda-weighted arrows from inefficient units to their targets, or projection arrows to the frontier.

  • plot_io_parcoo() and plot_io_radar() gain a variables argument to show a subset of the inputs/outputs: "all" (default), "inputs", "outputs", or a vector of variable names or positions.

  • New dataset taiwanese_banks: a balanced panel of 22 Taiwanese commercial banks over 2009-2011 (Kao & Liu, 2014, ), used as the worked example for plot_panel_io_biplot().

  • plot_panel_io_biplot() now accepts inputs, outputs, id and period as integer column positions as well as names, and repels the loading-vector labels (via ggrepel) so they no longer overlap.

  • New x_angle argument on plot_io_parcoo(), plot_io_heatmap(), plot_cem_heatmap(), plot_cem_weights_heatmap() and plot_io_distributions() to rotate long x-axis tick labels for readability.

  • A single DMU passed to labels now fades the rest of the plot into a focus view (keeping the chosen DMU’s sub-network / trajectory); the fade argument tunes the fade level or disables it.

deaviz 0.1.0

CRAN release: 2026-07-02

  • First release: a from-scratch rebuild of the DEA-Viz visualization methods as an R package.
  • All functions are built on a single validated dea_data object and follow a compute_* / plot_* naming convention.
  • Analysis: compute_efficiency(), compute_cross_efficiency(), compute_cross_efficiency_weights(), standardize_weights(), compute_multiplier_weights(), compute_som().
  • Visualization: plot_efficiency_bar(), plot_efficiency_comparison(), plot_cross_efficiency(), plot_cem_unfolding(), plot_cross_efficiency_weights_heatmap(), plot_pca_biplot(), plot_mds(), plot_porembski_network(), plot_panel_biplot(), plot_som(), plot_som_components(), plot_costa_frontier(), plot_parallel_coordinates(), plot_3d_scatter(), plot_histograms(), plot_dotplots(), plot_pairwise_scatter(), plot_input_output_heatmap().
  • Ships the chinese_cities example dataset (35 cities, 3 inputs, 3 outputs).
  • Consistent, colour-blind-safe visual style across all plots: Okabe-Ito qualitative palette, viridis sequential palette, and a shared minimal theme.