r - 在ggplot中按组堆叠发散条形图

Question

我正在尝试做一个这样的图表：

我们的想法是绘制 3 个数量，在这个混合堆叠条形图中，我们有一个数据框，其中一行用于负值，两行用于正值，但是我需要将负数与第一个正数条堆叠在一起，我还需要3种颜色。我到目前为止的代码如下：（数据框已经具有所需的形状）：

df3 <- read.table(
text = 
"region group metric somevalue
blue T1 epsilon 63
blue T2 epsilon -40
red T1 epsilon 100
blue T1 kappa 19
blue T2 kappa -30
red T1 kappa 75
blue T1 zulu  50
blue T2 zulu -18
red T1 zulu 68", header=TRUE)

p2 <- ggplot(df3, aes(x = metric, y = somevalue, fill=region))+
  geom_col(aes(fill = group), width = 0.7) + geom_bar(position = 'dodge', stat='identity')
p2

请帮帮我，如果您认为必须修改数据框，请告诉我。谢谢

标签： rggplot2charts

Answer 1

堆叠和躲避总是有点棘手。在您的情况下，可以这样实现：

1. 转换region为factor. （这确保第 3 步有效）
2. 将您的数据集一分为二以获得负值和正值。
3. 使用 tidy::complete 填充数据集，以便每个数据集包含度量、区域和组的“所有”组合。（这确保躲避工作
4. 使用两个 geom_col 层来绘制正值和负值position="dodge"。我添加na.rm = TRUE以删除我们通过添加的缺失值complete。

library(ggplot2)
library(dplyr)
library(tidyr)

df3$region <- factor(df3$region)

df3_neg <- filter(df3, somevalue < 0) %>% 
  tidyr::complete(region, group, metric)

df3_pos<- filter(df3, somevalue > 0) %>% 
  tidyr::complete(region, group, metric)

p2 <- ggplot(df3, aes(somevalue, metric)) +
  geom_col(aes(alpha = group, fill=region), data = df3_pos, position = "dodge", na.rm = TRUE) +
  geom_col(aes(alpha = group, fill=region), data = df3_neg, position = "dodge", na.rm = TRUE) +
  scale_fill_identity() +
  scale_alpha_manual(values = c(T2 = .6, T1 = 1)) +
  guides(alpha = FALSE)
p2

编辑添加注释可以以相同的方式实现，例如我下面的代码使用两个geom_text在我使用的栏旁边添加值，position_dodge2(.9)以便标签与栏很好地对齐：

p2 <- ggplot(df3, aes(somevalue, metric)) +
  geom_col(aes(alpha = group, fill=region), data = df3_pos, position = "dodge", na.rm = TRUE) +
  geom_col(aes(alpha = group, fill=region), data = df3_neg, position = "dodge", na.rm = TRUE) +
  geom_text(aes(x = somevalue + 1, label = somevalue), data = df3_pos, position = position_dodge2(width  = .9), hjust = 0, na.rm = TRUE) +
  geom_text(aes(x = somevalue - 1, label = somevalue), data = df3_neg, , position = position_dodge2(width  = .9), hjust = 1, na.rm = TRUE) +
  scale_fill_identity() +
  scale_alpha_manual(values = c(T2 = .6, T1 = 1)) +
  guides(alpha = FALSE)
p2

EDIT2添加表确实是另一回事。在那种情况下，我会选择patchwork这意味着制作绘图来模仿表格布局。为了使躲避工作或确保表格行与您为每个表格列绘制的条形图对齐。基本方法可能如下所示：

library(patchwork)

# 1. Make a dataframe with all combinations of region and metric using expand_grid
d_table <- expand_grid(region = unique(df3$region), metric = unique(df3$metric))

# 2. Add columns with the table content
d_table$column1 <- LETTERS[1:6]
d_table$column2 <- letters[1:6]

# 3. Make a plot for each column of the table
p_column1 <- ggplot(d_table, aes(y = metric, x = 1, label = column1)) +
  geom_text(aes(group = region), position = position_dodge2(width = .9), na.rm = TRUE) +
  scale_x_continuous(position = "top", breaks = 1, labels = "column1") +
  labs(y = NULL, x = "") +
  theme(axis.text.y = element_blank(), 
        axis.text.x.bottom = element_blank(),
        axis.ticks = element_blank(), 
        plot.margin = unit(rep(0, 4), "pt"), 
        panel.background = element_rect(fill = NA))
p_column2 <- ggplot(d_table, aes(y = metric, x = 1, label = column2)) +
  geom_text(aes(group = region), position = position_dodge2(width = .9), na.rm = TRUE) +
  scale_x_continuous(position = "top", breaks = 1, labels = "column2") +
  labs(y = NULL, x = "") +
  theme(axis.text.y = element_blank(), 
        axis.text.x.bottom = element_blank(),
        axis.ticks = element_blank(), 
        plot.margin = unit(rep(0, 4), "pt"), 
        panel.background = element_rect(fill = NA))
# 4. Add the table columns via patchwork
p2 + p_column1 + p_column2 + plot_layout(widths = c(1, .1, .1))