ggplot2中geom_errorbar异常与条件设置点形状失效问题

我有一组拆分存储在列表中的数据框，需为每个数据框绘制折线图：未检出（低于检出限）的数据使用空心符号（点形状值为1）展示，检出数据（均值+标准差）使用填充符号（点形状值为16）展示。
以下为可复现问题的示例代码：

OutdoorAir <- data.frame(
  "Sample" = c("Air", "Air", "Air", "Air","Aerosol", "Aerosol", "Aerosol", "Aerosol"), 
  "Date" = c("1980", "1990", "2000", "2010", "1980", "1990", "2000", "2010"), 
  "Value" = c(331, 250, 484, 221, 545, 131, 417, 476),
  "Locality" = c("Paris", "Lyon", "Paris", "Lyon", "Paris", "Lyon", "Paris", "Lyon"),
  "Less.Than" = c("<","<","","<","","","<",""),
  "Uncertainty" = c(NA,NA,45,NA,10,8,NA,71),
  "Nuclides" = c("K-40","K-40", "K-40", "K-40", "C-14", "C-14", "C-14", "C-14"))

OutdoorAir_l <- split(OutdoorAir,list(OutdoorAir$Sample,OutdoorAir$Locality, OutdoorAir$Nuclides), drop=TRUE)
lapply(names(OutdoorAir_l), function(i){
  ggplot(OutdoorAir_l[[i]], aes(Date, Value, group = Nuclides, col = as.factor(Nuclides))) +
    colorization +
    geom_errorbar(aes(ymin=Value-Uncertainty, ymax=Value+Uncertainty), color = "black") +
    geom_point() +
    scale_shape_manual(values=ifelse("Less.Than" == "<",c(1),c(15)), guide = "none") +
    ggsave(paste(i, ".png", sep = ""),  dpi = 100, width = 30, height = 20, units = "cm")
})
dev.off()

数据规则说明 #

• 若检测结果低于检出限，Less.Than列填充值为<，此时Value列存储的实际为检出限值，需在图中显示为空心圆
• 其余情况Value列为检测均值，Uncertainty列为标准差

1. 上述代码中scale_shape_manual函数无法按Less.Than列的条件正确匹配点形状，写法存在什么错误？
2. geom_errorbar绘制的误差线存在样式不一致问题：部分误差线为带两端水平端帽的标准样式，部分仅显示单条竖线，该问题的成因是什么？

1. 点形状匹配错误的原因 #

代码中形状映射存在3个核心问题：

• 未在美学映射中绑定形状字段：geom_point()没有将shape参数和Less.Than列关联，ggplot无法根据每行数据的属性动态分配点形状
• 条件判断逻辑完全失效：ifelse("Less.Than" == "<",c(1),c(15))是在判断字符串常量"Less.Than"是否等于<，而非逐行判断数据框中Less.Than列的值，永远只会返回单一值15
• 形状值不符合需求：代码中检出数据配置的形状是15（实心方块），和需求要求的16（实心圆）不匹配

2. 误差线样式不一致的原因 #

异常样式的核心成因是低于检出限的数据Uncertainty列为NA：

• 当Uncertainty为有效值时，ymin = Value - Uncertainty和ymax = Value + Uncertainty是两个不同数值，误差线有正常的上下延伸长度，会完整绘制两端的水平端帽
• 当Uncertainty为NA时，ymin和ymax计算结果均为NA，ggplot会默认在数据点位置绘制长度为0的误差线，两端端帽完全重合，视觉上就呈现为单条竖线
• 低于检出限的数据本身不需要展示误差线，直接过滤这部分数据的误差线绘制即可解决样式问题。

修正后可运行代码 #

library(ggplot2)
# 注：原代码中colorization为自定义主题/色板，此处可替换为自己的配色配置
OutdoorAir_l <- split(OutdoorAir,list(OutdoorAir$Sample,OutdoorAir$Locality, OutdoorAir$Nuclides), drop=TRUE)
lapply(names(OutdoorAir_l), function(i){
  p <- ggplot(OutdoorAir_l[[i]], aes(x = Date, y = Value, group = Nuclides, col = as.factor(Nuclides))) +
    # 仅为检出数据绘制误差线，过滤未检出的NA值
    geom_errorbar(data = ~subset(.x, Less.Than != "<"), 
                  aes(ymin=Value-Uncertainty, ymax=Value+Uncertainty), 
                  color = "black", width = 0.2) +
    # 补充折线图所需的连线层
    geom_line() +
    # 将形状映射到Less.Than字段
    geom_point(aes(shape = Less.Than), size = 3) +
    # 手动匹配两类点的形状：< 对应空心圆(1)，正常值对应实心圆(16)
    scale_shape_manual(values = c("<" = 1, "" = 16), guide = "none")
  ggsave(paste0(i, ".png"), plot = p, dpi = 100, width = 30, height = 20, units = "cm")
})
dev.off()

内容的提问来源于stack exchange，提问作者Sylvain