在R中处理含NA的ggalluvial格式数据并绘制桑基图求助

核心思路 #

ggalluvial对数据格式和NA处理要求严格，要实现按角色（A/B/C/D）统一着色的5阶段桑基图，需确保数据结构正确映射路径、处理NA值，并将颜色绑定到角色类型而非阶段。

步骤1：数据准备与清洗 #

假设你的原始数据是宽格式（每行对应一条角色流转路径，含5个阶段列和频率列），先处理NA值（避免路径断裂）：

library(ggplot2)
library(ggalluvial)
library(dplyr)

# 示例宽格式数据（替换为你的真实数据）
role_data <- tibble(
  Stage1 = c("A", "B", "A", NA),
  Stage2 = c("B", NA, "A", "C"),
  Stage3 = c(NA, "A", "B", "D"),
  Stage4 = c("C", "A", "C", NA),
  Stage5 = c("D", "C", "D", "B"),
  Frequency = c(10, 15, 8, 5)
)

# 将NA替换为显性类别（如"无角色"），保留流转路径的完整性
role_data_clean <- role_data %>%
  mutate(across(starts_with("Stage"), ~ifelse(is.na(.), "无角色", .)))

如果你的数据是长格式，需确保每条路径有唯一ID关联各阶段记录：

# 示例长格式数据（替换为你的真实数据）
role_long_data <- tibble(
  PathID = rep(1:4, each = 5),
  Stage = rep(paste0("阶段", 1:5), 4),
  Role = c("A", "B", "无角色", "C", "D",
           "B", "无角色", "A", "A", "C",
           "A", "A", "B", "C", "D",
           "无角色", "C", "D", "无角色", "B"),
  Frequency = rep(c(10,15,8,5), each =5)
)

步骤2：绘制桑基图 #

宽格式数据绘图代码

ggplot(role_data_clean,
       aes(y = Frequency,
           axis1 = Stage1, axis2 = Stage2, axis3 = Stage3, axis4 = Stage4, axis5 = Stage5)) +
  # 绘制冲积流，按角色统一着色
  geom_alluvium(aes(fill = Stage1), width = 1/12) +
  # 绘制分层节点，白色填充加黑色边框
  geom_stratum(width = 1/12, fill = "white", color = "black") +
  # 添加分层节点标签
  geom_text(stat = "stratum", aes(label = after_stat(stratum))) +
  # 设置x轴阶段顺序与标签
  scale_x_discrete(limits = c("Stage1", "Stage2", "Stage3", "Stage4", "Stage5"),
                   expand = c(0.05, 0.05),
                   labels = paste0("阶段", 1:5)) +
  # 自定义角色颜色（可根据示例图调整）
  scale_fill_manual(values = c(
    "A" = "#E74C3C", "B" = "#2ECC71", 
    "C" = "#3498DB", "D" = "#9B59B6", 
    "无角色" = "#BDC3C7"
  )) +
  labs(title = "5阶段角色流转桑基图", x = "角色阶段", y = "组合频率") +
  theme_minimal()

长格式数据绘图代码

ggplot(role_long_data,
       aes(x = Stage, stratum = Role, alluvium = PathID,
           y = Frequency, fill = Role)) +
  geom_alluvium(width = 1/12) +
  geom_stratum(width = 1/12, color = "black") +
  geom_text(stat = "stratum", aes(label = Role)) +
  # 自定义角色颜色（可根据示例图调整）
  scale_fill_manual(values = c(
    "A" = "#E74C3C", "B" = "#2ECC71", 
    "C" = "#3498DB", "D" = "#9B59B6", 
    "无角色" = "#BDC3C7"
  )) +
  labs(x = "角色阶段", y = "组合频率", title = "5阶段角色流转桑基图") +
  theme_minimal()

关键注意事项 #

• NA处理：必须将NA替换为显性类别（如"无角色"），否则ggalluvial会自动忽略含NA的路径，导致流转逻辑缺失。
• 着色逻辑：通过fill = Role（长格式）或fill = Stage1（宽格式，需确保路径起始角色能代表核心角色）实现同角色同色。
• 格式匹配：宽格式需保证每行是完整路径，长格式需用PathID关联同一路径的不同阶段记录，频率值需对应整个路径的频次。
• 样式调整：通过width参数修改冲积流和分层节点的宽度，scale_x_discrete调整x轴标签和间距，匹配你的示例图样式。

内容的提问来源于stack exchange，提问作者Anakin Skywalker