Overview

This document was generated on October 31, 2026. It demonstrates how to transform multi-dimensional categorical data into a continuous visual stream using a Sankey diagram rendered via plotly in R.

Sankey diagrams excel at illustrating proportional shifts across categorical stages. Here, we track passenger demographics and survival outcomes from the RMS Titanic disaster of April 1912.

Dataset Description

The analysis uses the built-in R Titanic dataset, derived from the official passenger manifest and historical records compiled after the sinking.

Structure & Dimensions

The dataset is a 4-dimensional cross-tabulated array consisting of 2,201 records categorized across four factors:

  • Class: Passenger booking class or crew designation (1st, 2nd, 3rd, Crew).
  • Sex: Biological sex (Male, Female).
  • Age: Age category (Child, Adult).
  • Survived: Final survival status (No, Yes).

Data Summary

The table below aggregates the total passenger counts across passenger class, sex, and final survival outcomes:

library(dplyr)
library(knitr)

# Convert 4D array to data frame
titanic_df <- as.data.frame(Titanic) %>%
  filter(Freq > 0)

# Summary table grouped by Class, Sex, and Survived
summary_table <- titanic_df %>%
  group_by(Class, Sex, Survived) %>%
  summarise(Passengers = sum(Freq), .groups = "drop") %>%
  arrange(Class, Sex, Survived)

kable(summary_table, caption = "Titanic Manifest Summary by Class, Sex, and Outcome")
Titanic Manifest Summary by Class, Sex, and Outcome
Class Sex Survived Passengers
1st Male No 118
1st Male Yes 62
1st Female No 4
1st Female Yes 141
2nd Male No 154
2nd Male Yes 25
2nd Female No 13
2nd Female Yes 93
3rd Male No 422
3rd Male Yes 88
3rd Female No 106
3rd Female Yes 90
Crew Male No 670
Crew Male Yes 192
Crew Female No 3
Crew Female Yes 20

Interactive Flow Diagram

The Sankey diagram maps categorical transitions left-to-right along a two-stage flow:

\[\text{Passenger Class} \longrightarrow \text{Sex} \longrightarrow \text{Survival Outcome}\]

Hover over nodes or connecting flows to inspect exact passenger volumes.

library(plotly)

# Stage 1: Class -> Sex
flow_class_sex <- titanic_df %>%
  group_by(Class, Sex) %>%
  summarise(value = sum(Freq), .groups = "drop") %>%
  rename(source_name = Class, target_name = Sex)

# Stage 2: Sex -> Survived
flow_sex_survived <- titanic_df %>%
  group_by(Sex, Survived) %>%
  summarise(value = sum(Freq), .groups = "drop") %>%
  rename(source_name = Sex, target_name = Survived) %>%
  mutate(target_name = paste("Outcome:", target_name))

# Combine stage flows
all_flows <- bind_rows(flow_class_sex, flow_sex_survived)

# Map nodes to 0-indexed vectors for Plotly
unique_nodes <- c("1st", "2nd", "3rd", "Crew", 
                  "Male", "Female", 
                  "Outcome: No", "Outcome: Yes")

node_lookup <- setNames(seq(0, length(unique_nodes) - 1), unique_nodes)

all_flows <- all_flows %>%
  mutate(
    source = node_lookup[source_name],
    target = node_lookup[target_name]
  )

# Define node palette
node_colors <- c(
  "#2b5c8f", "#3690c0", "#67a9cf", "#a6bddb", # Class nodes
  "#8856a7", "#e08214",                      # Sex nodes
  "#d73027", "#1a9850"                       # Survival outcome nodes
)

# Render Sankey Widget
plot_ly(
  type = "sankey",
  orientation = "h",
  node = list(
    label = unique_nodes,
    color = node_colors,
    pad = 18,
    thickness = 22,
    line = list(color = "#333333", width = 0.5)
  ),
  link = list(
    source = all_flows$source,
    target = all_flows$target,
    value  = all_flows$value
  )
) %>%
  layout(
    title = list(text = "Titanic Passenger Flows: Class → Sex → Survival", font = list(size = 16)),
    font = list(size = 12)
  )

Analytical Observations

  1. Class Distribution: Crew members and 3rd-class passengers represented the largest initial node blocks in the manifest.
  2. Gender Ratio: Male passengers heavily outnumbered female passengers, particularly among crew members and 3rd-class travelers.
  3. Outcome Disparities: The path width leading into Outcome: Yes illustrates high survival proportions among female passengers, highlighting the historical enforcement of the “women and children first” protocol during evacuation.