Star Wars Edges Scatter Plot Explorer

An interactive scatter plot of the Star Wars character interaction dataset. Each point is one edge record (a source → target interaction within an episode).

Dataset Overview

• Filename: data/starwars_edges.csv (loaded at /data/starwars_edges.csv)
• Rows: 1013 (as reported by the project)
• Columns: 4
• Description: Edge list of interactions between Star Wars characters, with the episode in which the interaction occurs and a weight-like value.

Column Descriptions

Meanings of value are inferred from the column name and typical edge-list conventions; they are not confirmed by documentation in the dataset.

Visualization Mappings

• X axis: episode — the episode is the only ordered discrete identifier with a meaningful numeric order, so it provides a natural horizontal position.
• Y axis: value — the only other quantitative field, giving each edge a vertical magnitude.
• Color: source — the character initiating the interaction. It is nominal with many categories; the color legend lists them and hovering a legend entry fades the other characters. target and episode are available from the Color menu as alternatives.

Because the dataset has only two quantitative fields, the scatter plot's X and Y menus offer episode and value. Categorical fields are offered only as color encodings.

Interaction

• Hover a point to see its source, target, episode, and value.
• Hover a legend entry to highlight that category.
• Use the menus to remap axes or color.
• Press V to show Voronoi cell borders.

Visualization Ideas

1. Episode-by-episode interaction volume

   • Main question: How does the number of recorded character interactions vary across episodes, helping the social analytics audience compare activity by episode?
   • Relevant data fields: episode (count edge rows).
   • Suggested visual encoding: A bar chart with episode on the x-axis and the number of edge records on the y-axis.
   • Interaction or annotation: Hover a bar to show its episode and exact edge count.
   • Implementation note: Load /data/starwars_edges.csv with d3.csv() and count rows by parsed episode; this measures records, not unique relationships.

2. Total interaction value by episode

   • Main question: How does the recorded magnitude of interactions change across episodes?
   • Relevant data fields: episode, value.
   • Suggested visual encoding: A column chart of summed value for each episode, with episode on the x-axis.
   • Interaction or annotation: Add a toggle between sum and mean so viewers can distinguish aggregate magnitude from average edge value.
   • Implementation note: Parse numeric fields from the loaded CSV and aggregate value by episode; describe the measure as recorded value because its meaning is undocumented.

3. Distribution of edge values by episode

   • Main question: Are interaction values typically similar across episodes, or do some episodes contain more extreme records?
   • Relevant data fields: episode, value.
   • Suggested visual encoding: A box plot for each episode, showing median, quartiles, and range of value.
   • Interaction or annotation: Hover a box to reveal its record count and summary statistics.
   • Implementation note: Compute summaries from finite value records grouped by episode; do not interpret the values beyond their recorded magnitude.

4. Interaction value over episode, with record-level detail

   • Main question: Which individual interaction records have relatively high or low values, and where do they occur in the episode sequence?
   • Relevant data fields: episode, value, source, target.
   • Suggested visual encoding: Retain the current scatter plot: episode on x, value on y, and character source as color.
   • Interaction or annotation: Use the existing Voronoi hover tooltip to show source → target, episode, and value; let viewers switch color to target or episode.
   • Implementation note: Use the existing data-driven scatter plot and encodings; avoid adding character fields as numeric axes.

5. Most frequent sources

   • Main question: Which characters most often appear as the initiating source in the recorded interactions?
   • Relevant data fields: source (count edge rows).
   • Suggested visual encoding: A sorted horizontal bar chart of records per source character.
   • Interaction or annotation: Selecting a character highlights that source’s records and shows the corresponding episode and target details.
   • Brief implementation note: Count loaded CSV rows by source; label results as source-record frequency, not as a measure of overall character importance.

6. Most frequent targets

   • Main question: Which characters most often appear as the receiving target?
   • Relevant data fields: target (count edge rows).
   • Suggested visual encoding: A sorted horizontal bar chart, distinct from the source ranking, of records per target.
   • Interaction or annotation: Link each bar to a filtered list or scatter plot of its incoming edge records.
   • Brief implementation note: Group rows by target and count them; keep source and target roles explicit because they are not interchangeable fields.

7. Character role comparison

   • Main question: For each character, how do their observed initiating and receiving record counts compare?
   • Relevant data fields: source, target (count edge rows for each role).
   • Suggested visual encoding: A dumbbell plot with one position for source count and one for target count per character.
   • Interaction or annotation: Sort by either role count and highlight a selected character’s records in both roles.
   • Brief implementation note: Derive the character list from observed source and target values, then compute the two counts from the edge list; do not call these measures of influence.

8. Most common directed character pairs

   • Main question: Which exact source → target pairs recur most often in the records?
   • Relevant data fields: source, target (count matching edge rows).
   • Suggested visual encoding: A ranked horizontal bar chart of directed pairs, with bar length representing record count.
   • Interaction or annotation: Selecting a pair filters the detail view to its records by episode and value.
   • Brief implementation note: Group by both source and target without collapsing direction; derive all pairs from the loaded CSV.

9. Character-to-character interaction matrix

   • Main question: Which character pairs have recorded edges, and how does the recorded value or count vary between them?
   • Relevant data fields: source, target, value.
   • Suggested visual encoding: A matrix with source characters as rows and target characters as columns; use cell color for edge count and allow switching to summed value.
   • Interaction or annotation: Hover a cell to list its episode records; provide character search to navigate a large matrix.
   • Brief implementation note: Build row and column categories from observed names. Clearly label whether the cell shows count or summed recorded value.

10. Bipartite source–target network

    • Main question: How are initiating and receiving roles connected across the dataset?
    • Relevant data fields: source, target, value, episode.
    • Suggested visual encoding: A bipartite network with source-role nodes on one side and target-role nodes on the other; edge width represents record count or summed value.
    • Interaction or annotation: Hover an edge to show its pair and records; filter by episode.
    • Brief implementation note: Keep the two role columns distinct in the layout, even when a character name occurs in both roles; derive edges and totals from the CSV.

11. Force-directed character network

    • Main question: What overall pattern of recorded connections emerges when characters and their interactions are viewed together?
    • Relevant data fields: source, target, value, episode.
    • Suggested visual encoding: A force-directed network with character nodes and directed edges; use arrows for source-to-target direction and optionally scale width by edge count.
    • Interaction or annotation: Hover a node to highlight its incident edges; select an episode to show only records from that episode.
    • Brief implementation note: Construct nodes from observed names and links from the edge list. State that the layout reflects the chosen algorithm, not a measured social distance.

12. Episode-filtered network small multiples

    • Main question: How does the pattern of recorded character-to-character links differ from one episode to another?
    • Relevant data fields: episode, source, target, value.
    • Suggested visual encoding: Small-multiple directed networks, one panel per episode, using consistent node identities and edge encoding.
    • Interaction or annotation: Hover a link to display its source, target, episode, and value.
    • Brief implementation note: Split loaded records by episode; use consistent scales where possible and avoid implying a node or edge is absent from the story merely because it is absent from that episode’s records.

13. Episode filter for record exploration

    • Main question: Which source–target records and values appear in the episode a viewer wants to examine?
    • Relevant data fields: episode, source, target, value.
    • Suggested visual encoding: An episode selector linked to the current episode-by-value scatter plot.
    • Interaction or annotation: Selecting an episode filters points and updates a count of visible records; retain tooltips for pair details.
    • Brief implementation note: Filter the loaded rows by episode and use the existing scatter plot scales and tooltip behavior.

14. Episode-by-character activity heatmap

    • Main question: Which characters appear most frequently in records in each episode, either as source or as target?
    • Relevant data fields: episode, source, target (count edge-row appearances by role).
    • Suggested visual encoding: A heatmap with episodes on one axis and characters on the other; color shows record appearances, with a role switch for source versus target.
    • Interaction or annotation: Hover a cell to show its count; clicking it filters to the matching episode and role.
    • Brief implementation note: Derive character names and episode values from the data and count one selected role at a time so each cell has a clear definition.

15. Episode sequence of a selected directed pair

    • Main question: In which episodes is a particular source → target pair recorded, and what values are attached to those records?
    • Relevant data fields: source, target, episode, value.
    • Suggested visual encoding: A selected-pair dot plot with episode on the x-axis and value on the y-axis.
    • Interaction or annotation: Choose the pair from observed names; use point tooltips to reveal the record details.
    • Brief implementation note: Filter the CSV to the selected ordered pair. Preserve multiple records rather than assuming one record per pair and episode.

16. Value concentration and outliers

    • Main question: Are recorded values concentrated in a small portion of edge records, or spread more evenly?
    • Relevant data fields: value, source, target, episode.
    • Suggested visual encoding: A sorted dot plot of individual records by value, with points colored by episode or source.
    • Interaction or annotation: Hover high- and low-value points to inspect the corresponding pair and episode; offer an episode filter.
    • Brief implementation note: Sort finite values for display and retain row-level tooltips. Avoid labeling records as causal or significant based on value alone.

17. Source–target flow diagram

    • Main question: Which source and target groups account for the largest recorded flows between character roles?
    • Relevant data fields: source, target, value.
    • Suggested visual encoding: A Sankey-style diagram with source nodes on the left, target nodes on the right, and link width representing edge count or summed recorded value.
    • Interaction or annotation: Hover a link to show the pair and selected aggregation; switch between count and sum.
    • Brief implementation note: Build the flow from observed directed pairs and explicitly label the link-width metric; do not treat value as a known causal strength.

18. Episode-by-pair interaction timeline

    • Main question: How are recorded directed pairs distributed across the episode sequence?
    • Relevant data fields: episode, source, target, value.
    • Suggested visual encoding: A timeline with episode on the x-axis and one row per selected directed pair; mark each recorded edge, with size or color encoding value.
    • Interaction or annotation: Search or select pairs from the dataset and show pair details on hover.
    • Brief implementation note: Generate the timeline from actual edge records and support pair filtering to keep the display readable; do not infer interactions in episodes without a row.

19. Value versus record frequency for directed pairs

    • Main question: Do pairs with more records also have higher summed recorded values?
    • Relevant data fields: source, target, value (derive record count and summed value per pair).
    • Suggested visual encoding: A scatter plot with pair record count on x and summed value on y; use labels or color for selected directed pairs.
    • Interaction or annotation: Hover a point to show its source → target, record count, and value sum; allow filtering to one episode.
    • Brief implementation note: Compute pair-level aggregates from the loaded rows and label both axes as derived summaries, with no assumption about the underlying meaning of value.

20. Linked overview and detail dashboard

    • Main question: How can the audience move from overall episode patterns to the specific character records behind them?
    • Relevant data fields: episode, source, target, value.
    • Suggested visual encoding: Combine episode-level interaction counts, a value-by-episode scatter plot, and a compact source–target pair ranking in coordinated views.
    • Interaction or annotation: Brushing an episode range or selecting a character or pair filters the other views; a reset control restores all loaded records.
    • Brief implementation note: Derive every panel from the same CSV rows and keep selections synchronized; label the record-count and recorded-value measures distinctly.

References

Inspired by:

• Interactive Scatter Plot — Visualization nk2fwdG0qh1d5PXbpwNhr

The reference implementation's structure (context providers, D3-in-React render functions, Voronoi hit-testing, and the horizontal color legend) was adapted to this dataset. The reference's own dataset and data-specific columns were not reused.