ENA and ONA with RS.data

This article walks through fitting a standard ENA (undirected) model and an ONA (directed) model from the built-in RS.data dataset, then rendering both as live interactive plots with qeviz.

Data and model setup

library(qeviz)
library(tma)
library(rENA)

data(RS.data)

units_by <- c("Condition", "UserName")
codes    <- c(
  "Data", "Technical.Constraints", "Performance.Parameters",
  "Client.and.Consultant.Requests", "Design.Reasoning", "Collaboration"
)

hoo_rules <- tma::conversation_rules(
  (Condition %in% UNIT$Condition) & (GroupName %in% UNIT$GroupName)
)

context_model <- tma::contexts(
  RS.data,
  units_by  = make.names(units_by),
  hoo_rules = hoo_rules
)

multidim_arr <- tma::context_tensor(
  RS.data,
  sender_cols    = NULL,
  receiver_cols  = NULL,
  default_window = 8,
  default_weight = 1
)

result_ena <- tma::accumulate(
  tensor        = multidim_arr,
  codes         = codes,
  context_model = context_model,
  ordered       = FALSE
)

result_ona <- tma::accumulate(
  tensor        = multidim_arr,
  codes         = codes,
  context_model = context_model,
  ordered       = TRUE
)

ena_set <- rENA::model(result_ena)
ona_set <- rENA::model(result_ona, directed = TRUE)

Creating a plot object

qe_plot() wraps a set object in a qe_plot that can be passed through the chain API. No rendering happens yet — the plot is fully lazy.

qe_ena <- qe_plot(ena_set)
qe_ona <- qe_plot(ona_set)

Inspect what groups and units are available:

qe_groups(qe_ena)
#> [1] "FirstGame"  "SecondGame"
head(qe_units(qe_ena))
#> [1] "FirstGame::steven z"    "FirstGame::akash v"     "FirstGame::alexander b"
#> [4] "FirstGame::brandon l"   "FirstGame::christian x" "FirstGame::jordan l"

Standard ENA (undirected)

Show both group mean networks overlaid. Each chain call returns a new qe_plot — the original is unchanged.

p <- qe_ena |>
  qe_group() |>                                  # show all group means + CI
  qe_edges(qe_groups(qe_ena)[1],
           also = qe_groups(qe_ena)[2])          # overlay both networks

p

Compare groups by subtraction — positive weights favour the first group, negative weights favour the second:

p_sub <- qe_ena |>
  qe_group() |>
  qe_edges(qe_groups(qe_ena)[1],
           compare = qe_groups(qe_ena)[2])

p_sub

ONA (directed)

Directed edges are rendered as tapered arrows whose apex points toward the stronger directional contributor.

p_ona <- qe_ona |>
  qe_group() |>
  qe_edges(qe_groups(qe_ona)[1], also = qe_groups(qe_ona)[2])

p_ona

Individual unit networks

Use qe_units() to list unit IDs, then pass one to qe_edges(unit = ...). The group mean markers stay visible so the individual network can be read in context.

unit_id <- qe_units(qe_ena)[1]
cat("Plotting unit:", unit_id, "\n")
#> Plotting unit: FirstGame::steven z

p_unit <- qe_ena |>
  qe_group() |>             # show all group means for context
  qe_edges(unit = unit_id)  # individual edge network

p_unit

Adding unit points

qe_points() overlays the individual unit scatter dots:

p_pts <- qe_ena |>
  qe_group() |>
  qe_edges(qe_groups(qe_ena)[1]) |>
  qe_points()

p_pts

Label control

qe_labels() controls label visibility for nodes, group means, and unit points independently. Accepted values: "on", "off", "click", "auto".

p_labels <- qe_ena |>
  qe_group() |>
  qe_edges(qe_groups(qe_ena)[1]) |>
  qe_points() |>
  qe_labels(nodes = "auto", means = "on", points = "click")

p_labels

Explicit data and plot settings (qeviz 0.5)

Networks, points and means can also come straight from data, as rENA’s ena.plot.network() / ena.plot.points() / ena.plot.group() take them. Explicit layers are drawn on the model’s scales, so widths and colours stay comparable with the name-based plots above.

g1 <- as.matrix(ena_set$line.weights[ena_set$line.weights$Condition == qe_groups(qe_ena)[1]])
first_half <- ena_set$points[ena_set$points$GMR1 > 0]   # means rotation: GMR1 is dimension 1

qe_plot(ena_set, title = "Units with GMR1 > 0", group_col = "Condition",
        dims = c("GMR1", "SVD2"), range = "network") |>
  qe_edges(weights = colMeans(g1), threshold = 0.05) |>
  qe_points(points = first_half, color = "#555555") |>
  qe_group(points = first_half, label = "GMR1 > 0", intervals = "crosshairs") |>
  qe_labels(nodes = "on", font_size = 13)

Exporting standalone files

qe_export_html() writes a self-contained HTML file — no server or internet connection required:

qe_export_html(qe_ena |> qe_group() |> qe_edges(qe_groups(qe_ena)[1]),
               "ena_plot.html")

qe_export_html(qe_ona |> qe_group() |> qe_edges(qe_groups(qe_ona)[1]),
               "ona_plot.html")

Branching from a common base works naturally since each chain call copies the plot:

base <- qe_ena |> qe_group()

qe_export_html(qe_edges(base, qe_groups(qe_ena)[1]), "group1.html")
qe_export_html(qe_edges(base, qe_groups(qe_ena)[2]), "group2.html")
qe_export_html(qe_edges(base, unit = qe_units(qe_ena)[1]), "unit1.html")