如何绘制含缺失值(NA)的柱状图？Comtrade数据可视化问题

问题描述 #

使用comtradr包从联合国Comtrade数据库拉取数据并可视化，完成数据抓取与重塑后，印度、美国的部分数据行存在NA值，导致这两个国家的条形从柱状图中消失。尝试在geom_bar参数中添加na.rm=TRUE未解决问题，还收到两条「Removed 2 rows containing missing values (position_stack)」的警告。

原因分析 #

问题出在数据聚合阶段而非绘图阶段：aggregate()函数遇到分组内的NA值时，会直接返回该分组的NA结果，最终生成的df_import2和df_export2中，印度、美国的Count字段为NA。ggplot绘制时会自动移除含NA的行，而geom_bar中的na.rm=TRUE仅处理绘图层的NA，无法解决源头数据的NA问题。

解决方案 #

在数据聚合时直接忽略分组内的NA值，通过mean(..., na.rm=TRUE)计算有效平均值。同时用tidyverse语法简化代码逻辑，具体修改如下：

修改后的完整代码 #

# Notes.------------------------------------------------------------------------

options(scipen=999)

# Packages.---------------------------------------------------------------------

library(tidyverse)
library(comtradr)
library(patchwork)

# Extract export data with query.-----------------------------------------------

df_a <- ct_search(
  reporters = c("China", "USA", "India", "Germany"),
  partners = "World",
  trade_direction = trade_list,
  freq = "annual",
  commod_codes = "2501",
  type = "goods"
)

df_b <- ct_search(
  reporters = c("Australia", "Canada", "Chile", "Mexico"),
  partners = "World",
  trade_direction = trade_list,
  freq = "annual",
  commod_codes = "2501",
  type = "goods"
)

df_c <- ct_search(
  reporters = c("Russian Federation", "Brazil"),
  partners = "World",
  trade_direction = trade_list,
  freq = "annual",
  commod_codes = "2501",
  type = "goods"
)

# Select relevant columns & join datasets.---------------------------------------

df <- bind_rows(df_a, df_b, df_c) %>%
  select(year, qty, reporter, trade_flow)

# 数据聚合时忽略NA值，计算有效平均值
df_import2 <- df %>% 
  filter(trade_flow == "Import") %>% 
  group_by(reporter) %>%
  summarise(Count = mean(qty, na.rm = TRUE) / 1000) %>%
  rename(Country = reporter)

df_export2 <- df %>% 
  filter(trade_flow == "Export") %>% 
  group_by(reporter) %>%
  summarise(Count = mean(qty, na.rm = TRUE) / 1000) %>%
  rename(Country = reporter)

# Build plots from extracted data.----------------------------------------------

import_plot <- ggplot(data = df_import2, aes(x = reorder(Country, Count), y = Count)) +
  geom_bar(stat = "identity", fill = "lightblue") +
  coord_flip() +
  theme_bw()

export_plot <- ggplot(data = df_export2, aes(x = reorder(Country, Count), y = Count)) +
  geom_bar(stat = "identity", fill = "#de4048") +
  coord_flip() +
  theme_bw()

import_plot / export_plot

关键修改说明 #

1. 在summarise()计算平均值时加入na.rm=TRUE，确保分组内存在NA值时仍能生成有效均值，避免返回NA结果。
2. 用group_by()+summarise()替代aggregate()，代码更简洁易读，符合tidyverse工作流。
3. 绘图时无需再添加na.rm=TRUE，因为数据已无NA值，警告会自动消失，印度、美国的条形将正常显示。

内容的提问来源于stack exchange，提问作者Jay Bee