geom_col绘制柱形图y值异常问题求助

背景信息 #

• 分析目标：探究反应时（rt_mean）与组内分类变量Assignment（Yourself/Stranger）的关系，每个被试对应多条rt_mean记录
• 数据示例：

participant Assignment rt_mean
         <dbl> <chr>        <dbl>
 1     7148832 Yourself      601.
 2     7148832 Yourself      554.
 3     7148832 Yourself      604.
 4     7148832 Stranger      810.
 5     7148832 Stranger      821.
 6     7148832 Stranger      712.
 7     7148833 Yourself      651.
 8     7148833 Yourself      725.
 9     7148833 Yourself      784.
10     7148833 Stranger      693.

• 手动计算均值：mean_yourself = 686.2，mean_stranger = 749.3
• 问题现象：用geom_col()绘制的柱形图中，Yourself柱形高于Stranger（与计算结果相反）且数值偏高，但geom_boxplot()显示正常

问题原因 #

geom_col()的默认行为是将同一x分组下的所有y值堆叠求和，而非计算均值。从数据示例看，Yourself的记录条数（6条）多于Stranger（4条），即使单条均值更低，总和依然更大，导致柱形高度与手动计算的均值趋势完全相反。

解决办法 #

方法1：先汇总均值再绘图（推荐，逻辑清晰） #

先通过分组统计得到每个Assignment的均值，再基于汇总后的数据绘图：

# 分组计算均值
reaction_time_summary <- reaction_time %>%
  dplyr::group_by(Assignment) %>%
  dplyr::summarize(rt_mean_avg = mean(rt_mean, na.rm = TRUE))

# 绘制柱形图
reaction_time_plot <- reaction_time_summary %>%
  ggplot2::ggplot(aes(x = Assignment, y = rt_mean_avg, fill = Assignment)) +
  geom_col(position = position_dodge()) +
  labs(x = "Assignment", y = "Reaction Time in ms") +
  theme_bw() +
  scale_fill_manual(values = c("#ABB0B8", "#4C4E52"))
reaction_time_plot

方法2：直接在绘图时指定统计变换 #

使用stat_summary()替代geom_col()，直接在图层中指定计算均值：

reaction_time_plot <- reaction_time %>%
  ggplot2::ggplot(aes(x = Assignment, y = rt_mean, fill = Assignment)) +
  stat_summary(fun = mean, geom = "col", position = position_dodge()) +
  labs(x = "Assignment", y = "Reaction Time in ms") +
  theme_bw() +
  scale_fill_manual(values = c("#ABB0B8", "#4C4E52"))
reaction_time_plot

内容的提问来源于stack exchange，提问作者JSK