Plot backends: plotly and qeviz

rENA draws plots with one of two backends. plotly is the default. qeviz (>= 0.5.0, from https://cran.qe-libs.org; a suggested package) draws the same plots with qeviz, the library behind webENA's network plots, and returns a qeviz htmlwidget.

Choosing a backend

Per plot, with ena.plot(set, backend = "qeviz"), or for the session with options(rENA.plot.backend = "qeviz"). ena.plotter and ena.plot.subtraction take backend too and pass it to every plot they create.

Existing plotting code works unchanged on either backend: ena.plot.network, ena.plot.points, ena.plot.group, and the pipe API (add_network(), add_points(), add_group(), with_means(), check_range(), show()) add layers to whichever backend the plot was created with.

Side by side

The same rENA code on RS.data drawn with each backend (left and middle), and the same view written with qeviz's own functions (right). The code for every panel is in PLOT-BACKENDS.md in the package sources, also at https://qe-libs.org/cran/pkg/rENA/PLOT-BACKENDS.html.

One group's mean network and its mean, from ena.plot.network() and ena.plot.group(). On qeviz, line width is absolute per weight and each node's size is the sum of its edge widths:

A subtraction, FirstGame - SecondGame, with both means. Colour and width are on the model-wide scale, so the small differences in a subtraction draw lighter than a group's network; multiplier (or ena.plotter()'s multipliers) magnifies them:

Every unit and both means with scale.to = "points", from ena.plot.points() and ena.plot.group():

What differs on qeviz

• $plot is a qeviz htmlwidget, not a plotly object, so code that edits it with plotly::layout() or other plotly functions does not apply. print() shows it as before.

• Scaling. Line width is absolute per weight, colour intensity is scaled across the whole model, and a node's size is the sum of its edge widths. Separate plots of one model are therefore directly comparable. thickness, opacity, saturation, scale.range, scale.weights and node.size are ignored.

• Multipliers. ena.plot(multiplier =), add_network(edge.multiplier =), and the ena.plotter() and ena.plot.subtraction() multipliers magnify the drawn network rather than multiplying the weights. ena.plotter() and ena.plot.subtraction() label their plots "(scaled Nx)".

• Point size. point.size is halved to match qeviz's units, so the default sizes look the same.

• Means. Each ena.plot.group() call keeps its own confidence and outlier intervals. An outlier interval drawn as "crosshairs" is drawn as a box. method must be "mean".

• Ranges are symmetric: scale.to is "network", "points" or a number; list(x =, y =) is an error.

• Default palette is Paul Tol's "bright" (qeviz::qe_palette()). Colours you pass are used as given.

Not available on qeviz

Trajectories and movies (ena.plot.trajectory, add_trajectory(), ena.plot.movie, with_trajectory()) warn with class rENA_qeviz_unsupported and return the plot unchanged; use backend = "plotly" for them. Dashed or dotted edges, label offsets, legends and per-point confidence or outlier intervals are not drawn.

Arguments that would change the picture but cannot be honoured warn once per session. Plotly-only layout options (showticklabels, ticks, axispadding, autosize, automargin, displayModeBar) are ignored without a warning.

See also

ena.plot, ena.plotter, ena.plot.interactive

Examples

if (requireNamespace("qeviz", quietly = TRUE) &&
    utils::packageVersion("qeviz") >= "0.5.0") {
  data(RS.data)
  codeNames <- c("Data", "Technical.Constraints", "Performance.Parameters",
    "Client.and.Consultant.Requests", "Design.Reasoning", "Collaboration")
  accum <- ena.accumulate.data(
    units = RS.data[, c("Condition", "UserName")],
    conversation = RS.data[, c("Condition", "GroupName")],
    codes = RS.data[, codeNames],
    window.size.back = 4
  )
  set <- ena.make.set(accum)
  first <- set$meta.data$Condition == "FirstGame"

  plot <- ena.plot(set, title = "FirstGame", backend = "qeviz")
  plot <- ena.plot.network(plot, network = colMeans(as.matrix(set$line.weights[first, ])))
  plot <- ena.plot.group(plot, as.matrix(set$points[first, ]), labels = "FirstGame",
                         confidence.interval = "box")
  print(plot)
}