移除分组条形图中的NA类别（R语言ggplot2实现）

问题场景 #

处理包含250列的调研数据，通过循环遍历所有列生成以sector为x轴、回答占比为y轴、各数据列为填充色的分组条形图，但图表中出现了透明的NA类别条形，尝试scale_fill_discrete(na.translate = FALSE)（仅移除图例NA）、fill = subset(data,!is.na(data[[variable]]))、ggplot(data=na.omit(data[[variable]]))均无效。

示例数据：

q1 <- factor(c("yes",NA,"no","yes",NA,"yes","no","yes"))
q2 <- factor(c("Albania","USA","Albania","Albania","UK",NA,"UK","Albania"))
q3 <- factor(c(0,1,NA,0,1,1,NA,0))
q4 <- factor(c(0,NA,NA,NA,1,NA,0,0))
q5 <- factor(c("Dont know","Prefer not to answer","Agree","Disagree",NA,"Agree","Agree",NA))
q6 <- factor(c(1,NA,3,5,800,NA,900,2))
sector <- factor(c("Energy","Water","Energy","Other","Other","Water","Transportation","Energy"))

data <- data.frame(q1,q2,q3,q4,q5,q6,sector)

原绘图代码：

by_sector <- lapply(names(data), function(variable) {
  ggplot(
    data = data,
    mapping = aes(x=sector,fill = data[[variable]])
  ) +
    geom_bar(aes( y=..count../tapply(..count.., ..x.. ,sum)[..x..]), position="dodge") +
    labs(x = variable, y = "% of total", fill = "Response", caption = paste("Total =", sum(!is.na(data[[variable]])))) +
    geom_text(aes( y=..count../tapply(..count.., ..x.. ,sum)[..x..], label=scales::percent(..count../tapply(..count.., ..x.. ,sum)[..x..],accuracy = 0.1) ),
              stat="count", position=position_dodge(1), vjust=0.5)+
    #scale_fill_brewer(palette = "Accent")+
    scale_fill_discrete(na.translate = FALSE) +
    theme_bw() +
    theme(panel.grid.major.y = element_blank()) +
    coord_flip()
  
})

解决方案 #

关键修改是在ggplot()的data参数中，针对当前遍历的变量过滤掉该列含NA的行，而非直接使用完整数据集。修改后的代码如下：

by_sector <- lapply(names(data), function(variable) {
  # 过滤当前变量列的NA值，保留其他列完整数据
  filtered_data <- data[!is.na(data[[variable]]), ]
  
  ggplot(
    data = filtered_data,
    mapping = aes(x = sector, fill = .data[[variable]])
  ) +
    geom_bar(aes(y = ..count../tapply(..count.., ..x.. , sum)[..x..]), position = "dodge") +
    labs(x = variable, y = "% of total", fill = "Response", 
         caption = paste("Total =", nrow(filtered_data))) +
    geom_text(aes(y = ..count../tapply(..count.., ..x.. , sum)[..x..], 
                  label = scales::percent(..count../tapply(..count.., ..x.. , sum)[..x..], accuracy = 0.1)),
              stat = "count", position = position_dodge(1), vjust = 0.5) +
    scale_fill_discrete(na.translate = FALSE) +
    theme_bw() +
    theme(panel.grid.major.y = element_blank()) +
    coord_flip()
})

说明 #

1. 过滤数据逻辑：通过data[!is.na(data[[variable]]), ]仅移除当前变量列含NA的行，绘图时不会包含NA类别数据，自然不会生成NA条形。
2. 规范列访问方式：用.data[[variable]]替代data[[variable]]，这是ggplot推荐的编程式列访问方法，避免潜在的评估问题。
3. 更新统计标注：用nrow(filtered_data)替代原sum(!is.na(data[[variable]]))，结果一致但更直观。

之前尝试的方法无效原因：

• scale_fill_discrete(na.translate = FALSE)仅控制图例是否显示NA，不会移除绘图数据中的NA；
• fill = subset(data,!is.na(data[[variable]]))错误地将填充映射到数据框子集，而非当前列的有效值；
• ggplot(data=na.omit(data[[variable]]))传入的是向量而非完整数据框，ggplot无法识别sector等其他变量。

内容的提问来源于stack exchange，提问作者dryl