The Phantom Network: Darth Jar Jar’s Guide to Social Network Analysis in R

MSBA 580 // Week 9 // Classified Holonet Transmission

THE PHANTOM
NETWORK

A Star Wars field guide to nodes, edges, aliases, hidden influence, and the machinery of rebellion.

Prepared by Kilgore Trout // Educational fan-fiction analysis

Mission briefing

Networks are made of two ingredients:

Vertices (nodes): the people, organizations, planets, products, or pages.
Edges (links): the relationships connecting those vertices.

For marketers, network analysis can identify influential customers, tightly connected communities, likely referrals, and paths through which information may spread.

For this fan-fiction version of the lab, you are Darth Jar Jar: publicly a clumsy representative from Naboo, secretly the architect of a galaxy-spanning influence network. The Star Wars framing changes the labels, not the mathematics. The same graph methods apply to customers, social-media accounts, websites, organizations, and supply chains.

Reproducibility check

This table records the package versions used to knit the report. If results ever change after an update, this is the first place to check.

package_versions <- tibble(
  package = required_packages,
  version = vapply(
    required_packages,
    function(pkg) as.character(utils::packageVersion(pkg)),
    character(1)
  )
)

render_table(
  package_versions,
  caption = "Package versions used for this analysis",
  digits = 3
)

Package versions used for this analysis
package	version
tidyverse	2.0.0
igraph	2.3.3
knitr	1.51
kableExtra	1.4.1
scales	1.4.0
networkD3	0.4.1
htmlwidgets	1.6.4

First, the basics: Darth Jar Jar’s inner circle

Why the original lines 86–88 needed special attention

The original lab created the starter graph in one long graph_from_literal() statement and then immediately plotted it. That can work, but it is difficult to audit and easy to break with a misspelled name such as Farther.

This version replaces the fragile one-line graph with:

A clearly labeled vertex table.
A clearly labeled edge table.
Validation checks before plotting.
A fixed random seed for a repeatable layout.

# -------------------------------------------------------------------------
# STARTER VERTEX TABLE
# -------------------------------------------------------------------------
# Each row represents one character in Darth Jar Jar's hidden network.
darth_vertices <- tribble(
  ~name,               ~faction,          ~public_role,
  "Darth Jar Jar",     "Shadow Sith",     "Naboo representative",
  "Darth Sidious",     "Shadow Sith",     "Supreme Chancellor",
  "Count Dooku",       "Separatists",      "Separatist leader",
  "General Grievous",  "Separatists",      "Military commander",
  "Queen Amidala",     "Naboo",            "Naboo leader",
  "Boss Nass",         "Gungan Council",   "Gungan leader",
  "Anakin Skywalker",  "Jedi",             "Jedi Knight",
  "Obi-Wan Kenobi",    "Jedi",             "Jedi Master"
)

# -------------------------------------------------------------------------
# STARTER EDGE TABLE
# -------------------------------------------------------------------------
# Each row represents one relationship. An explicit edge table is easier to
# inspect than a dense graph literal and is the preferred pattern when a
# network will later receive attributes.
darth_edges <- tribble(
  ~from,               ~to,                ~relationship,
  "Darth Jar Jar",     "Darth Sidious",    "secret alliance",
  "Darth Jar Jar",     "Count Dooku",      "covert coordination",
  "Darth Sidious",     "Count Dooku",      "Sith command",
  "Darth Jar Jar",     "Queen Amidala",    "public trust",
  "Darth Jar Jar",     "Boss Nass",        "Gungan influence",
  "Queen Amidala",     "Boss Nass",        "Naboo alliance",
  "Darth Jar Jar",     "Anakin Skywalker", "subtle manipulation",
  "Darth Sidious",     "Anakin Skywalker", "apprentice grooming",
  "Anakin Skywalker",  "Obi-Wan Kenobi",   "Jedi mentorship",
  "Count Dooku",       "General Grievous", "military command",
  "Darth Sidious",     "General Grievous", "strategic control"
)

# Build an undirected graph because these starter relationships are being
# treated as mutual connections for introductory network analysis.
g5 <- igraph::graph_from_data_frame(
  d = darth_edges,
  directed = FALSE,
  vertices = darth_vertices
)

# Fail early with an informative error if the graph was built incorrectly.
stopifnot(
  igraph::vcount(g5) == nrow(darth_vertices),
  igraph::ecount(g5) == nrow(darth_edges),
  igraph::is_simple(g5)
)

# Degree is the number of direct connections for each character.
starter_degree <- igraph::degree(g5, mode = "all")

# Force-directed layouts include randomness. Setting a seed makes the same
# graph appear in the same approximate position each time the report knits.
set.seed(580)
starter_layout <- igraph::layout_with_fr(g5)

# Map node size to degree so more-connected characters look larger.
starter_size <- scales::rescale(
  starter_degree,
  to = c(24, 48)
)

# Create one repeatable color for each faction.
faction_levels <- sort(unique(igraph::V(g5)$faction))
faction_colors <- setNames(
  scales::hue_pal()(length(faction_levels)),
  faction_levels
)

plot(
  g5,
  layout = starter_layout,
  vertex.size = starter_size,
  vertex.color = faction_colors[igraph::V(g5)$faction],
  vertex.frame.color = "white",
  vertex.label = stringr::str_wrap(igraph::V(g5)$name, width = 14),
  vertex.label.color = "black",
  vertex.label.cex = 0.85,
  edge.width = 2,
  edge.color = "grey65",
  main = "Darth Jar Jar's Inner Circle\nNode size = number of direct connections"
)

legend(
  "topleft",
  legend = names(faction_colors),
  col = faction_colors,
  pch = 19,
  pt.cex = 1.4,
  bty = "n",
  title = "Faction"
)

The starter network demonstrates several foundational concepts:

Chain: Obi-Wan connects to Anakin, who connects to Darth Jar Jar.
Triangle: Darth Jar Jar, Darth Sidious, and Count Dooku are mutually connected.
Multiple paths: Darth Jar Jar can reach General Grievous through either Count Dooku or Darth Sidious.
Degree: Darth Jar Jar has the most direct connections and is therefore the most locally connected character in this small graph.

starter_degree_table <- tibble(
  character = names(starter_degree),
  degree = as.integer(starter_degree)
) |>
  arrange(desc(degree), character)

render_table(
  starter_degree_table,
  caption = "Direct connections in Darth Jar Jar's starter network",
  digits = 0
)

Direct connections in Darth Jar Jar’s starter network
character	degree
Darth Jar Jar	5
Darth Sidious	4
Anakin Skywalker	3
Count Dooku	3
Boss Nass	2
General Grievous	2
Queen Amidala	2
Obi-Wan Kenobi	1

Understanding Checkpoint

The simulated Holonet dataset

The original lab modeled Facebook users and friendships. Here we simulate a smaller Holonet influence network with the same structure:

Vertices: individual accounts with descriptive attributes.
Edges: undirected social or communication connections.
Network shape: preferential attachment, where already-connected accounts are more likely to attract new connections.

This preserves the original marketing lesson: social networks usually contain a small number of hubs and many less-connected users.

Step 1: Simulate the vertex table

set.seed(580)

n_users <- 150

known_aliases <- c(
  "Darth Jar Jar", "Darth Sidious", "Queen Amidala", "Boss Nass",
  "Count Dooku", "General Grievous", "Anakin Skywalker", "Obi-Wan Kenobi",
  "Ahsoka Tano", "Mace Windu", "Yoda", "Bail Organa", "Mon Mothma",
  "Nute Gunray", "Wat Tambor", "Jango Fett", "Boba Fett", "Cad Bane",
  "R2-D2", "C-3PO"
)

vertices <- tibble(
  # id must be unique because graph_from_data_frame() uses it as the
  # symbolic vertex name.
  id = seq_len(n_users),

  # Use recognizable aliases first, then reproducible generic accounts.
  alias = if_else(
    id <= length(known_aliases),
    known_aliases[id],
    sprintf("Holonet Agent %03d", id)
  ),

  # Attributes support later joins and segmentation exercises.
  birthday = sample(
    seq.Date(
      from = as.Date("1985-01-01"),
      to = as.Date("2005-12-31"),
      by = "day"
    ),
    size = n_users,
    replace = TRUE
  ),

  homeworld = sample(
    c(
      "Naboo", "Coruscant", "Tatooine", "Mandalore",
      "Kashyyyk", "Kamino", "Ryloth", "Corellia"
    ),
    size = n_users,
    replace = TRUE
  ),

  faction_id = sample(
    c(
      "Galactic Republic", "Jedi Order", "Separatists",
      "Gungan Council", "Crime Syndicate", "Independent"
    ),
    size = n_users,
    replace = TRUE
  ),

  academy_id = sample(
    c(
      "Jedi Temple", "Royal Academy", "Gungan Grand Army",
      "Mandalorian Training", "Trade Federation", "None"
    ),
    size = n_users,
    replace = TRUE
  )
)

# Derive the birth month after the base table is created.
# Birth month is more useful than an exact date in this small simulation
# because it creates enough matches to demonstrate the join reliably.
vertices <- vertices |>
  mutate(
    birth_month_number = as.integer(format(birthday, "%m")),
    birth_month = factor(
      month.abb[birth_month_number],
      levels = month.abb
    )
  )

render_table(
  vertices |>
    select(id, alias, birthday, birth_month, homeworld, faction_id, academy_id) |>
    slice_head(n = 8),
  caption = "First 8 Holonet accounts in the simulated vertex table",
  digits = 0
)

First 8 Holonet accounts in the simulated vertex table
id	alias	birthday	birth_month	homeworld	faction_id	academy_id
1	Darth Jar Jar	1989-08-29	Aug	Coruscant	Gungan Council	Jedi Temple
2	Darth Sidious	1991-01-04	Jan	Coruscant	Gungan Council	None
3	Queen Amidala	1998-04-19	Apr	Naboo	Galactic Republic	Mandalorian Training
4	Boss Nass	1986-02-27	Feb	Corellia	Jedi Order	Royal Academy
5	Count Dooku	1988-04-09	Apr	Corellia	Crime Syndicate	Jedi Temple
6	General Grievous	1998-07-07	Jul	Kashyyyk	Independent	Royal Academy
7	Anakin Skywalker	1987-09-12	Sep	Kashyyyk	Separatists	None
8	Obi-Wan Kenobi	1986-01-13	Jan	Naboo	Gungan Council	Royal Academy

Step 2: Simulate the edge table

set.seed(580)

# Preferential attachment generates a realistic hub-and-spoke structure.
g_sim <- igraph::sample_pa(
  n = n_users,
  power = 1.1,
  m = 3,
  directed = FALSE
)

# Remove any loops or duplicate edges as a defensive reproducibility step.
g_sim <- igraph::simplify(
  g_sim,
  remove.multiple = TRUE,
  remove.loops = TRUE
)

edges <- igraph::as_data_frame(g_sim, what = "edges") |>
  transmute(
    src = as.integer(from),
    dst = as.integer(to)
  ) |>
  arrange(src, dst)

stopifnot(
  all(edges$src %in% vertices$id),
  all(edges$dst %in% vertices$id),
  !any(edges$src == edges$dst)
)

render_table(
  edges |> slice_head(n = 8),
  caption = "First 8 simulated Holonet connections",
  digits = 0
)

First 8 simulated Holonet connections
src	dst
1	2
1	3
1	4
1	5
1	6
1	7
1	9
1	10

Understanding Checkpoint

Step 3: Build and validate the graph

fb_graph <- igraph::graph_from_data_frame(
  d = edges,
  vertices = vertices,
  directed = FALSE
)

# These checks confirm that no IDs were lost during graph construction.
stopifnot(
  igraph::vcount(fb_graph) == nrow(vertices),
  igraph::ecount(fb_graph) == nrow(edges),
  igraph::is_simple(fb_graph),
  igraph::is_connected(fb_graph)
)

fb_graph

## IGRAPH 30c5a9b UN-- 150 444 -- 
## + attr: name (v/c), alias (v/c), birthday (v/n), homeworld (v/c),
## | faction_id (v/c), academy_id (v/c), birth_month_number (v/n),
## | birth_month (v/x)
## + edges from 30c5a9b (vertex names):
##  [1] 1--2   1--3   1--4   1--5   1--6   1--7   1--9   1--10  1--11  1--12 
## [11] 1--19  1--20  1--21  1--25  1--26  1--28  1--29  1--33  1--47  1--50 
## [21] 1--57  1--72  1--77  1--79  1--87  1--89  1--96  1--97  1--101 1--102
## [31] 1--104 1--116 1--121 1--131 1--133 1--139 1--140 1--148 1--150 2--3  
## [41] 2--4   2--6   2--7   2--8   2--10  2--14  2--17  2--19  2--24  2--27 
## [51] 2--30  2--35  2--38  2--44  2--45  2--48  2--61  2--62  2--63  2--64 
## + ... omitted several edges

Inspect edge triplets

A triplet is an edge plus the attributes of both endpoints. This is useful when the question depends on who is connected and what those two accounts have in common.

edge_table <- graph_edges_integer(fb_graph)

vertex_a <- vertices |>
  rename_with(~ paste0("a_", .x))

vertex_b <- vertices |>
  rename_with(~ paste0("b_", .x))

triplets <- edge_table |>
  left_join(
    vertex_a,
    by = c("from" = "a_id")
  ) |>
  left_join(
    vertex_b,
    by = c("to" = "b_id")
  )

render_table(
  triplets |>
    select(
      from, a_alias, a_homeworld, a_faction_id,
      to, b_alias, b_homeworld, b_faction_id
    ) |>
    slice_head(n = 5),
  caption = "Sample triplets: one connection plus both accounts' attributes",
  digits = 0
)

Sample triplets: one connection plus both accounts’ attributes
from	a_alias	a_homeworld	a_faction_id	to	b_alias	b_homeworld	b_faction_id
1	Darth Jar Jar	Coruscant	Gungan Council	2	Darth Sidious	Coruscant	Gungan Council
1	Darth Jar Jar	Coruscant	Gungan Council	3	Queen Amidala	Naboo	Galactic Republic
1	Darth Jar Jar	Coruscant	Gungan Council	4	Boss Nass	Corellia	Jedi Order
1	Darth Jar Jar	Coruscant	Gungan Council	5	Count Dooku	Corellia	Crime Syndicate
1	Darth Jar Jar	Coruscant	Gungan Council	6	General Grievous	Kashyyyk	Independent

Step 4: Network algorithms and marketing applications

1. Find connected accounts with the same birth month

The original exercise searched for exact shared birthdays. In a 150-node simulation, exact matching dates among connected accounts can easily produce an empty table. Comparing birth month retains the personalization concept while creating a stable demonstration.

same_birth_month <- edge_table |>
  left_join(
    vertices |>
      select(id, alias, birth_month) |>
      rename(
        alias_a = alias,
        birth_month_a = birth_month
      ),
    by = c("from" = "id")
  ) |>
  left_join(
    vertices |>
      select(id, alias, birth_month) |>
      rename(
        alias_b = alias,
        birth_month_b = birth_month
      ),
    by = c("to" = "id")
  ) |>
  filter(birth_month_a == birth_month_b) |>
  transmute(
    account_a = alias_a,
    account_b = alias_b,
    shared_birth_month = birth_month_a
  ) |>
  arrange(shared_birth_month, account_a, account_b)

render_table(
  same_birth_month |> slice_head(n = 10),
  caption = "Connected Holonet accounts sharing a birth month",
  digits = 0
)

Connected Holonet accounts sharing a birth month
account_a	account_b	shared_birth_month
C-3PO	Holonet Agent 037	Jan
Darth Sidious	Holonet Agent 044	Jan
Darth Sidious	Holonet Agent 070	Jan
Darth Sidious	Obi-Wan Kenobi	Jan
Holonet Agent 044	Holonet Agent 086	Jan
Obi-Wan Kenobi	Holonet Agent 040	Jan
Obi-Wan Kenobi	Holonet Agent 091	Jan
Boss Nass	Wat Tambor	Feb
Holonet Agent 049	Holonet Agent 138	Feb
Holonet Agent 073	Holonet Agent 122	Feb

Marketing translation: shared attributes can support personalized promotions, community invitations, or audience segments.

2. Count triangles

A triangle is a set of three mutually connected accounts. Triangle-heavy users often sit inside cohesive communities where recommendations and information may spread efficiently.

triangle_counts <- tibble(
  id = as.integer(igraph::V(fb_graph)$name),
  triangle_count = as.integer(igraph::count_triangles(fb_graph))
) |>
  left_join(
    vertices |> select(id, alias, faction_id),
    by = "id"
  ) |>
  arrange(desc(triangle_count), alias)

render_table(
  triangle_counts |> slice_head(n = 10),
  caption = "Accounts participating in the most closed triangles",
  digits = 0
)

Accounts participating in the most closed triangles
id	triangle_count	alias	faction_id
1	50	Darth Jar Jar	Gungan Council
3	36	Queen Amidala	Galactic Republic
2	35	Darth Sidious	Gungan Council
6	29	General Grievous	Independent
4	15	Boss Nass	Jedi Order
11	14	Yoda	Separatists
10	13	Mace Windu	Crime Syndicate
21	9	Holonet Agent 021	Gungan Council
29	7	Holonet Agent 029	Crime Syndicate
16	7	Jango Fett	Independent

Understanding Checkpoint

3. Find friends of friends correctly in an undirected network

The original friend-of-friend pipeline converted every edge into a low-ID/high-ID pair and joined only one orientation. That misses valid two-step paths in an undirected graph.

The improved version first expands each edge into both directions, then:

Finds two accounts connected to the same mutual friend.
Removes self-pairs.
Standardizes pair order.
Removes pairs that are already directly connected.
Counts mutual friends.
Retains pairs that share an academy.

# Canonical direct-edge table used to remove already-connected pairs.
direct_edges <- edge_table |>
  transmute(
    a = pmin(from, to),
    b = pmax(from, to)
  ) |>
  distinct()

# Expand each undirected edge into two directed neighbor records.
adjacency <- bind_rows(
  edge_table |> transmute(account = from, mutual_friend = to),
  edge_table |> transmute(account = to, mutual_friend = from)
) |>
  distinct()

# Self-join through the shared mutual friend.
fof_candidates <- adjacency |>
  inner_join(
    adjacency,
    by = "mutual_friend",
    suffix = c("_a", "_b"),
    relationship = "many-to-many"
  ) |>
  transmute(
    a = pmin(account_a, account_b),
    b = pmax(account_a, account_b),
    mutual_friend = mutual_friend
  ) |>
  filter(a != b) |>
  distinct() |>
  anti_join(
    direct_edges,
    by = c("a", "b")
  ) |>
  count(
    a,
    b,
    name = "mutual_friend_count",
    sort = TRUE
  )

friends_of_friends <- fof_candidates |>
  left_join(
    vertices |>
      select(id, alias, academy_id) |>
      rename(
        alias_a = alias,
        academy_a = academy_id
      ),
    by = c("a" = "id")
  ) |>
  left_join(
    vertices |>
      select(id, alias, academy_id) |>
      rename(
        alias_b = alias,
        academy_b = academy_id
      ),
    by = c("b" = "id")
  ) |>
  filter(academy_a == academy_b) |>
  select(
    account_a = alias_a,
    account_b = alias_b,
    shared_academy = academy_a,
    mutual_friend_count
  ) |>
  arrange(desc(mutual_friend_count), account_a, account_b)

render_table(
  friends_of_friends |> slice_head(n = 10),
  caption = "Suggested introductions: mutual connections plus shared academy",
  digits = 0
)

Suggested introductions: mutual connections plus shared academy
account_a	account_b	shared_academy	mutual_friend_count
Boss Nass	General Grievous	Royal Academy	7
Queen Amidala	Holonet Agent 021	Mandalorian Training	5
Darth Jar Jar	Holonet Agent 044	Jedi Temple	4
Darth Sidious	Cad Bane	None	4
Darth Sidious	Holonet Agent 032	None	4
Darth Sidious	Holonet Agent 047	None	4
Mace Windu	Yoda	Mandalorian Training	4
Yoda	Holonet Agent 116	Mandalorian Training	4
Boss Nass	Holonet Agent 062	Royal Academy	3
Boss Nass	R2-D2	Royal Academy	3

Marketing translation: a recommendation is more credible when it combines network proximity with a relevant shared attribute.

4. Rank influential accounts with PageRank

Degree counts direct connections. PageRank also considers the importance of the accounts supplying those connections.

pagerank_vector <- igraph::page_rank(
  fb_graph,
  directed = FALSE,
  damping = 0.85
)$vector

influencers <- tibble(
  id = as.integer(names(pagerank_vector)),
  pagerank = as.numeric(pagerank_vector),
  degree = as.integer(igraph::degree(fb_graph))
) |>
  left_join(
    vertices |> select(id, alias, faction_id),
    by = "id"
  ) |>
  arrange(desc(pagerank), desc(degree), alias)

render_table(
  influencers |>
    select(alias, faction_id, degree, pagerank) |>
    slice_head(n = 10),
  caption = "Top Holonet accounts by PageRank",
  digits = 5
)

Top Holonet accounts by PageRank
alias	faction_id	degree	pagerank
Darth Jar Jar	Gungan Council	39	0.03853
Darth Sidious	Gungan Council	35	0.03499
Queen Amidala	Galactic Republic	28	0.02750
General Grievous	Independent	25	0.02438
Yoda	Separatists	23	0.02350
Holonet Agent 021	Gungan Council	17	0.01743
Mace Windu	Crime Syndicate	17	0.01724
Cad Bane	Jedi Order	16	0.01664
Boss Nass	Jedi Order	16	0.01609
R2-D2	Jedi Order	14	0.01515

Understanding Checkpoint

Section 5: Darth Jar Jar’s directed influence map

This section demonstrates PageRank in a directed network where arrows represent the flow of intelligence, pressure, or influence. The graph data support both the static contact map and the interactive D3 exploration that follows.

render_table(
  darth_rank |> slice_head(n = 10),
  caption = "Darth Jar Jar influence network ranked by weighted PageRank",
  digits = 5
)

Darth Jar Jar influence network ranked by weighted PageRank
node	faction	in_degree	out_degree	pagerank
Darth Jar Jar	Shadow Sith	5	8	0.26878
Galactic Senate	Galactic Institution	4	1	0.17897
Count Dooku	Separatists	3	2	0.09938
Jedi Council	Jedi	2	1	0.07829
General Grievous	Separatists	1	0	0.07027
Anakin Skywalker	Jedi	2	1	0.05406
Darth Sidious	Shadow Sith	1	3	0.05209
Gungan Council	Naboo	1	1	0.04517
Holonet News	Galactic Institution	1	1	0.04517
Queen Amidala	Naboo	1	2	0.04517

map_factions <- sort(unique(igraph::V(darth_graph)$faction))
map_colors <- setNames(
  scales::hue_pal()(length(map_factions)),
  map_factions
)

plot(
  darth_graph,
  layout = darth_layout,
  vertex.size = scales::rescale(
    igraph::V(darth_graph)$pagerank,
    to = c(22, 52)
  ),
  vertex.color = map_colors[igraph::V(darth_graph)$faction],
  vertex.frame.color = "white",
  vertex.label = stringr::str_wrap(igraph::V(darth_graph)$name, width = 15),
  vertex.label.cex = 0.78,
  vertex.label.color = "black",
  edge.width = scales::rescale(
    igraph::E(darth_graph)$weight,
    to = c(1, 5)
  ),
  edge.arrow.size = 0.45,
  edge.color = "grey55",
  main = paste(
    "Darth Jar Jar's Directed Influence Map",
    "Node size = weighted PageRank; arrow width = influence strength",
    sep = "\n"
  )
)

legend(
  "topleft",
  legend = names(map_colors),
  col = map_colors,
  pch = 19,
  pt.cex = 1.4,
  bty = "n",
  title = "Faction"
)

Interpretation: Darth Jar Jar appears important not merely because he sends many links, but because influential institutions and characters also direct information and access back toward him.

Section 6: Interactive D3 network map

networkD3 expects:

A node table indexed from zero.
A link table whose source and target columns use those zero-based indices.
Numeric node-size and link-value fields.

Darth Jar Jar’s Holonet Influence Map

nodes_d3 <- tibble(
  name = igraph::V(darth_graph)$name,
  faction = igraph::V(darth_graph)$faction,
  pagerank = as.numeric(igraph::V(darth_graph)$pagerank),

  # Precompute an easy-to-read radius range.
  node_size = scales::rescale(
    as.numeric(igraph::V(darth_graph)$pagerank),
    to = c(8, 24)
  )
)

links_d3 <- igraph::as_data_frame(darth_graph, what = "edges") |>
  transmute(
    # networkD3 requires zero-based indices rather than names.
    source = match(from, nodes_d3$name) - 1,
    target = match(to, nodes_d3$name) - 1,
    value = weight,
    channel = channel
  )

stopifnot(
  all(links_d3$source >= 0),
  all(links_d3$target >= 0),
  max(links_d3$source) < nrow(nodes_d3),
  max(links_d3$target) < nrow(nodes_d3)
)

darth_network_widget <- networkD3::forceNetwork(
  Links = as.data.frame(links_d3),
  Nodes = as.data.frame(nodes_d3),
  Source = "source",
  Target = "target",
  Value = "value",
  NodeID = "name",
  Group = "faction",
  Nodesize = "node_size",

  # Use htmlwidgets::JS explicitly. Calling JS() without its namespace can
  # fail when htmlwidgets has not been attached.
  radiusCalculation = htmlwidgets::JS("d.nodesize"),

  linkDistance = 165,
  charge = -500,
  opacity = 0.94,
  opacityNoHover = 1,
  zoom = TRUE,
  arrows = TRUE,
  fontSize = 12,
  fontFamily = "Trebuchet MS",
  bounded = TRUE,
  legend = TRUE,
  height = 700
)

htmlwidgets::onRender(
  darth_network_widget,
  "function(el, x) {
     var root = d3.select(el);

     root.selectAll('.nodetext')
       .style('fill', '#ffffff')
       .style('font-weight', '800')
       .style('paint-order', 'stroke')
       .style('stroke', '#07111f')
       .style('stroke-width', '3px')
       .style('stroke-linejoin', 'round')
       .style('opacity', 1);

     root.selectAll('.legend text')
       .style('fill', '#ffffff')
       .style('font-size', '12px')
       .style('font-weight', '800')
       .style('paint-order', 'stroke')
       .style('stroke', '#07111f')
       .style('stroke-width', '3px')
       .style('stroke-linejoin', 'round');

     root.selectAll('.legend rect')
       .style('stroke', '#edf3fb')
       .style('stroke-width', '1px');
   }"
)

Interactive interpretation: drag Darth Jar Jar away from the center. The network’s force layout pulls him back toward the institutions and characters with which he exchanges the strongest influence.

Understanding Checkpoint

Methodologies applied: decoding Luthen’s shadow network

Spoiler transmission: this section draws on both seasons of Andor and the political-spy network leading toward Rogue One.

This demonstration uses a deliberately small, classroom-scale evidence model. The relationships are anchored in official character and episode descriptions, but the calculated scores are analytic outputs from this selected model—not canonical power levels or definitive rankings. The objective is to show how an analyst moves from scattered observations to defensible conclusions.

The question we are trying to answer

How can a network reveal what kind of operative a character is—even when that character has few visible connections, multiple aliases, or deliberately hidden relationships?

The answer is not “count the most lines on screen.” We build the evidence chronologically, preserve direction and context, resolve identities, measure structure, remove key nodes, and only then translate the pattern into a conclusion.

The six-phase intelligence pipeline

Collect observable evidence

Record who communicates with whom, who issues orders, who supplies intelligence, who funds operations, and who is pursuing whom. Each observation becomes a candidate edge rather than an immediate conclusion.

Resolve identities before counting

Cassian Andor, Kassa, Clem, Keef Girgo, Tourist, and Varian Skye are different names used for the same character. Merge those names before calculating the network, or one highly connected character will be mistaken for several minor characters.

Define what every arrow means

An arrow may represent intelligence flow, command, political access, personal trust, coordination, or adversarial pursuit. Direction matters: Lonni feeding information to Luthen is analytically different from Luthen pressuring or handling Lonni.

Measure several kinds of importance

Degree measures visible contacts. PageRank rewards connections to important actors. Betweenness identifies brokers sitting on shortest paths. Two-step reach shows how far a message can travel quickly. No single metric is allowed to tell the whole story.

Stress-test the network

Remove Luthen, Lonni, Cassian, Saw, or Kleya one at a time. If communities split or routes disappear, the removed character was supplying structural cohesion, intelligence access, or operational redundancy.

Translate measurements into cautious claims

The final statement must separate canon from inference: ‘Lonni is canonically an ISB double agent’ is observed evidence; ‘Lonni behaves like a high-risk bridge whose loss isolates an intelligence channel’ is a network-derived conclusion.

One character, many names

Kassa → Cassian Andor → Clem → Keef Girgo → Tourist → Varian Skye

Character names merged before calculating centrality
character_name	story_context	network_treatment
Kassa	Kenari childhood	Count as Cassian Andor
Cassian Andor	Ferrix and the rebellion	Use as the combined analysis node
Clem	Aldhani mission	Count as Cassian Andor
Keef Girgo	Niamos and Narkina 5	Count as Cassian Andor
Tourist	Niamos arrest	Count as Cassian Andor
Varian Skye	Ghorman operation	Count as Cassian Andor

Why alias resolution changes the analytic story
model	nodes_used_for_one_person	highest_visible_degree	communities_visible_together	analytic_result
Aliases left fragmented	5	3	1	Centrality and reach are understated
Aliases resolved to Cassian Andor	1	10	6	Cross-community field role becomes visible

Observed The evidence model connects Lonni to Luthen, places Luthen in contact with Mon Mothma and Saw Gerrera, and combines Cassian Andor, Kassa, Clem, Keef Girgo, Tourist, and Varian Skye as one character node.

Derived Centrality, brokerage, reach, community position, and node-removal effects below are produced by this classroom network model.

Boundary A missing edge means “not included in this evidence model,” not “these characters never interacted.”

The contact map: what the structure makes visible

Read the map from the outside inward: local groups form recognizable clusters, while brokers occupy the connective tissue between clusters. Red arrows are pursuit, not cooperation; they are shown for situational awareness but excluded from collaboration-community calculations.

Interactive field console

Drag a node, zoom into a cluster, and watch the force layout pull brokers back toward the communities they connect. This interaction turns an abstract metric into a physical intuition: a structurally important node is difficult to move away from the network without stretching many relationships.

The metric panel

Selected structural signals from the Andor-era evidence model
character	role	direct_connections	two_step_reach	betweenness	pagerank	new_components_if_removed
Luthen Rael	strategist and handler	6	13	0.724	0.157	2
Cassian Andor	field operative	6	10	0.352	0.157	2
Lonni Jung	embedded intelligence source	2	8	0.343	0.051	1
Saw Gerrera	autonomous cell leader	3	8	0.133	0.080	1
Mon Mothma	political and financial bridge	2	7	0.133	0.061	1
Dedra Meero	counter-insurgency investigator	2	3	0.133	0.056	1
Kleya Marki	operations and continuity	2	9	0.000	0.057	0

Degree

Plain read: How many doors can this person open directly?

Expert read: Local connectivity; useful but vulnerable to visible-activity bias.

Betweenness

Plain read: How often must information pass through this person?

Expert read: Brokerage across shortest weighted paths; a proxy for coordination power and bottleneck risk.

PageRank

Plain read: Are the people connected to this person important too?

Expert read: Recursive prestige in a directed, weighted network.

Removal impact

Plain read: What breaks when this person disappears?

Expert read: Counterfactual fragmentation and resilience testing.

What we can responsibly conclude

Luthen Rael

Network inference

Plain-language conclusion: He is the switchboard. Political money, spies, field cells, and autonomous rebels connect through him.

Technical interpretation: High brokerage and removal impact identify a structural coordinator whose value comes from connecting otherwise separated communities.

Confidence: High within this selected model

Lonni Jung

Network inference

Plain-language conclusion: He does not need many friends to matter. He is the concealed wire carrying information out of the ISB.

Technical interpretation: Low-to-moderate degree can coexist with high bridge value. His position links an Imperial command environment to the Axis network, creating both intelligence value and single-channel risk.

Confidence: High for bridge interpretation

Cassian Andor

Network inference

Plain-language conclusion: Once his aliases are merged, he stops looking like several minor strangers and becomes a field connector spanning multiple worlds and cells.

Technical interpretation: Entity resolution recovers cross-community reach and prevents severe centrality attenuation. Alias fragmentation is a data-quality error, not merely a labeling inconvenience.

Confidence: High for identity-resolution lesson

Saw Gerrera

Network inference

Plain-language conclusion: Saw is powerful inside his own camp but deliberately difficult to absorb into someone else’s chain of command.

Technical interpretation: A locally cohesive partisan cluster with relatively few external bridges signals autonomy, modularity, and coordination friction rather than irrelevance.

Confidence: Moderate; depends on sampled ties

Kleya Marki

Network inference

Plain-language conclusion: Kleya is the continuity mechanism: the network can keep moving when Luthen cannot personally touch every operation.

Technical interpretation: Strong reciprocal ties to the hub and field operatives create operational redundancy and reduce hub overload.

Confidence: Moderate to high

Mon Mothma

Network inference

Plain-language conclusion: She connects respectable politics and resources to a covert network without resembling a battlefield hub.

Technical interpretation: Brokerage quality matters more than raw degree: a small number of cross-domain ties can supply disproportionate strategic value.

Confidence: High for bridge-type interpretation

Dedra Meero

Network inference

Plain-language conclusion: She is an adversarial mirror of the rebel network—assembling fragments until separate incidents appear to share a hidden center.

Technical interpretation: Her analytic task is link prediction and entity resolution under missing data. Adversarial edges describe detection pressure, not cooperative centrality.

Confidence: High as a methodological analogy

The expert-level lesson hidden inside the simple story

The most important character is not always the one with the most connections. A double agent can have low degree but enormous brokerage value. A political financier can have only a few edges yet connect domains that otherwise never touch. A field operative can appear statistically unimportant until aliases are resolved. A militant cell leader can be highly influential locally while remaining structurally distant from the larger coalition.

This is why serious network analysis combines entity resolution, edge semantics, direction, centrality, community structure, and counterfactual removal. The picture is not the conclusion; it is the interface through which the conclusion becomes understandable.

Canon anchors used for this demonstration

The official Star Wars Databank identifies Lonni Jung as an ISB supervisor secretly feeding intelligence to Luthen Rael.
Luthen’s official profile describes his contact with Mon Mothma, Saw Gerrera, Cassian, and a wider spy network.
Cassian Andor, Kassa, Clem, Keef Girgo, Tourist, and Varian Skye are combined into one character node before calculating centrality.
Combining the names prevents one character’s relationships from being split across several artificial nodes.

Practical marketing applications

Influencer identification: compare degree and PageRank rather than relying only on follower counts.
Community campaigns: use triangles and local clustering to find cohesive groups.
Recommendation systems: combine mutual connections with meaningful shared attributes.
Information diffusion: use directed graphs to understand who influences whom.
Interactive stakeholder tools: use D3 maps when decision-makers need to inspect specific accounts and relationships.

Key improvements made to the original lab

Replaced the fragile line-86 graph literal with explicit vertex and edge tables.
Corrected the Farther typo by removing the ambiguous family-network labels.
Added package checks without automatic installation.
Added package-version reporting and sessionInfo() for reproducibility.
Namespaced igraph conversion functions to avoid package conflicts.
Added graph validation with stopifnot().
Made force-directed layouts reproducible with set.seed().
Replaced HTML-inappropriate latex_options with Bootstrap table options.
Repaired the undirected friend-of-friend logic by expanding both edge directions.
Replaced sparse exact-birthday matching with a stable birth-month example.
Used htmlwidgets::JS() explicitly in the D3 graph.
Integrated Section 5 directly into the required project flow.
Added visible comments explaining purpose, assumptions, and output.

Final knowledge check

Understanding Checkpoint

Reproducibility record

sessionInfo()

## R version 4.6.1 (2026-06-24)
## Platform: x86_64-pc-linux-gnu
## Running under: Ubuntu 24.04.4 LTS
## 
## Matrix products: default
## BLAS:   /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3 
## LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.26.so;  LAPACK version 3.12.0
## 
## locale:
##  [1] LC_CTYPE=C.UTF-8       LC_NUMERIC=C           LC_TIME=C.UTF-8       
##  [4] LC_COLLATE=C.UTF-8     LC_MONETARY=C.UTF-8    LC_MESSAGES=C.UTF-8   
##  [7] LC_PAPER=C.UTF-8       LC_NAME=C              LC_ADDRESS=C          
## [10] LC_TELEPHONE=C         LC_MEASUREMENT=C.UTF-8 LC_IDENTIFICATION=C   
## 
## time zone: UTC
## tzcode source: system (glibc)
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
##  [1] htmlwidgets_1.6.4 networkD3_0.4.1   scales_1.4.0      kableExtra_1.4.1 
##  [5] igraph_2.3.3      lubridate_1.9.5   forcats_1.0.1     stringr_1.6.0    
##  [9] dplyr_1.2.1       purrr_1.2.2       readr_2.2.0       tidyr_1.3.2      
## [13] tibble_3.3.1      ggplot2_4.0.3     tidyverse_2.0.0  
## 
## loaded via a namespace (and not attached):
##  [1] sass_0.4.10        generics_0.1.4     xml2_1.6.0         stringi_1.8.7     
##  [5] hms_1.1.4          digest_0.6.39      magrittr_2.0.5     timechange_0.4.0  
##  [9] evaluate_1.0.5     grid_4.6.1         RColorBrewer_1.1-3 fastmap_1.2.0     
## [13] jsonlite_2.0.0     checkdown_0.0.13   viridisLite_0.4.3  textshaping_1.0.5 
## [17] jquerylib_0.1.4    cli_3.6.6          rlang_1.3.0        litedown_0.10     
## [21] commonmark_2.0.0   withr_3.0.3        cachem_1.1.0       yaml_2.3.12       
## [25] otel_0.2.0         tools_4.6.1        tzdb_0.5.0         vctrs_0.7.3       
## [29] R6_2.6.1           lifecycle_1.0.5    pkgconfig_2.0.3    bslib_0.11.0      
## [33] pillar_1.11.1      gtable_0.3.6       glue_1.8.1         systemfonts_1.3.2 
## [37] xfun_0.60          tidyselect_1.2.1   data.tree_1.2.0    rstudioapi_0.19.0 
## [41] knitr_1.51         farver_2.1.2       htmltools_0.5.9    rmarkdown_2.31    
## [45] svglite_2.2.2      compiler_4.6.1     S7_0.2.2           markdown_2.0

References

Csardi, G., and Nepusz, T. (2006). The igraph software package for complex network research. InterJournal, Complex Systems, 1695.
Page, L., Brin, S., Motwani, R., and Winograd, T. (1999). The PageRank Citation Ranking: Bringing Order to the Web. Stanford InfoLab.
Bostock, M., Ogievetsky, V., and Heer, J. (2011). D3: Data-Driven Documents. IEEE Transactions on Visualization and Computer Graphics.
igraph R documentation: https://r.igraph.org/reference/
networkD3 CRAN documentation: https://cran.r-project.org/package=networkD3
StarWars.com Databank. Luthen Rael: https://www.starwars.com/databank/luthen-rael
StarWars.com Databank. Lonni Jung: https://www.starwars.com/databank/lonni-jung
StarWars.com Databank. Cassian Andor: https://www.starwars.com/databank/cassian-andor
StarWars.com. “I Have Friends Everywhere” Episode Guide: https://www.starwars.com/series/andor/season-2-episode-5-episode-guide

The Phantom Network: Darth Jar Jar’s Guide to Social Network Analysis in R

Nodes, aliases, influence, and rebellion across the Holonet

Kilgore Trout — fan-fiction adaptation of the Week 9 lab by Jimmy Zhenning Xu, Ph.D.

2026-07-29

Mission briefing

Reproducibility check

First, the basics: Darth Jar Jar’s inner circle

Why the original lines 86–88 needed special attention

The simulated Holonet dataset

Step 1: Simulate the vertex table

Step 2: Simulate the edge table

Step 3: Build and validate the graph

Inspect edge triplets

Step 4: Network algorithms and marketing applications

1. Find connected accounts with the same birth month

2. Count triangles

3. Find friends of friends correctly in an undirected network

4. Rank influential accounts with PageRank

Section 5: Darth Jar Jar’s directed influence map

Section 6: Interactive D3 network map

Methodologies applied: decoding Luthen’s shadow network

The question we are trying to answer

The six-phase intelligence pipeline

Collect observable evidence

Resolve identities before counting

Define what every arrow means

Measure several kinds of importance

Stress-test the network

Translate measurements into cautious claims

One character, many names

The contact map: what the structure makes visible

Interactive field console

The metric panel

Degree

Betweenness

PageRank

Removal impact

What we can responsibly conclude

Luthen Rael

Lonni Jung

Cassian Andor

Saw Gerrera

Kleya Marki

Mon Mothma

Dedra Meero

The expert-level lesson hidden inside the simple story

Canon anchors used for this demonstration

Practical marketing applications

Key improvements made to the original lab

Final knowledge check

Reproducibility record

References