geom_alluvium双轴排序异常：仅单轴生效的解决求助

你的问题核心在于**lode.guidance = "forward"参数强制End轴的层（stratum）跟随Start轴的流向排序**，覆盖了你设置的End因子水平顺序，同时原始数据的因子转换存在类型匹配问题，以下是分步解决方法：

1. 修正因子转换逻辑 #

原始数据中Start和End是数值型，直接用字符型levels转换因子会导致类型隐式转换，建议先将列转为字符再设置因子顺序：

data$Start <- factor(as.character(data$Start), levels = c("2", "1", "3","4", "5"))
data$End <- factor(as.character(data$End), levels = c("3", "2", "1","4", "5","6"))

2. 覆盖End轴的排序引导 #

默认lode.guidance = "forward"会让End轴的层跟着Start轴的流向排列，需要通过以下两种方式之一修正：

方法一：调整lode.guidance参数 #

将lode.guidance设为"none"，让每个轴独立按自身因子顺序排序：

ggplot(data = data,
       aes(axis1 = Start, axis2 = End, y = Area)) +
  geom_alluvium(lode.guidance = "none",  # 关键修改：取消流向引导
                aes(fill = Start),width = 0.1,alpha = 0.8,knot.pos = 0,curve_type = "quintic") +
  geom_stratum(width = 1/7) + 
  geom_text(stat = "stratum", aes(label = after_stat(stratum))) +
  scale_x_discrete(limits = c("1960s", "2020s"), expand = c(0.15, 0.05)) +
  theme_void() +
  scale_fill_manual(values = c("1" = "#4a69bd",
                               "2" = "#54a0ff",
                               "3" = "#82ccdd",
                               "4" = "#b8e994",
                               "5" = "#ff9ff3",
                               "6" = '#e55039'))+
  labs(y = "", fill = "") +
  theme(legend.position = "none")

方法二：手动指定End轴的层排序 #

如果需要保留lode.guidance = "forward"的流线引导逻辑，可在geom_stratum和geom_text中通过order参数强制按End的因子顺序排序：

ggplot(data = data,
       aes(axis1 = Start, axis2 = End, y = Area)) +
  geom_alluvium(lode.guidance = "forward",
                aes(fill = Start),width = 0.1,alpha = 0.8,knot.pos = 0,curve_type = "quintic") +
  # 关键修改：给End轴的stratum指定排序
  geom_stratum(width = 1/7, aes(order = ifelse(after_stat(axis) == 2, after_stat(stratum), after_stat(order)))) + 
  geom_text(stat = "stratum", aes(label = after_stat(stratum), order = ifelse(after_stat(axis) == 2, after_stat(stratum), after_stat(order)))) +
  scale_x_discrete(limits = c("1960s", "2020s"), expand = c(0.15, 0.05)) +
  theme_void() +
  scale_fill_manual(values = c("1" = "#4a69bd",
                               "2" = "#54a0ff",
                               "3" = "#82ccdd",
                               "4" = "#b8e994",
                               "5" = "#ff9ff3",
                               "6" = '#e55039'))+
  labs(y = "", fill = "") +
  theme(legend.position = "none")

原理说明 #

• lode.guidance = "forward"会让下游轴（End）的层按上游轴（Start）的分组顺序排列，保证流线尽可能不交叉，但会覆盖你设置的因子顺序；
• 因子转换时必须保证原始值与levels的类型一致（数值转字符后再设因子），否则R会自动转换导致levels不匹配；
• 方法一适合完全独立排序双轴，方法二适合保留流线逻辑的同时强制End轴排序。

内容的提问来源于stack exchange，提问作者Evans Liu