R: grid.arrange 边缘图到 ggplot2 "heatmap"(geom_tile)

Question

我想在热图的顶部和右侧添加两个条形图，表示沿着热图表示的双变量分布的两个维度的边际分布。

这是一些代码:

library(gridExtra)
library(ggExtra)
library(cowplot)

# generate some data
df_hm = cbind(
  expand.grid(
    rows = sample(letters, 10), 
    cols = sample(LETTERS, 10)
  ), 
  value = rnorm(100)
)

# plot the heatmap
gg_hm = df_hm %>% 
  ggplot(aes(x = rows, y = cols, fill = value)) + 
  geom_tile() + 
  theme(legend.position = "bottom")

gg_rows = df_hm %>% 
  group_by(rows) %>% 
  summarize(value = mean(value)) %>% 
  ggplot(aes(x = rows,y = value)) + 
  geom_bar(stat = "identity", position = "dodge") 

gg_cols = df_hm %>% 
  group_by(cols) %>% 
  summarize(value = mean(value)) %>% 
  ggplot(aes(x = cols, y = value))+ 
  geom_bar(stat = "identity", position = "dodge") + 
  coord_flip()

gg_empty = df_hm %>% 
  ggplot(aes(x = cols, y = value)) +
  geom_blank() +
  theme(axis.text = element_blank(),
        axis.title = element_blank(),
        line = element_blank(),
        panel.background = element_blank())

# try this with grid.arrange
grid.arrange(gg_rows, gg_empty, gg_hm, gg_cols,
             ncol = 2, nrow = 2, widths = c(3, 1), heights = c(1, 3))

产生这个:

我想要做的是移动图形以对齐，如红色箭头所示:- (1, 1) 的 y 轴应与 (2, 1) 的 y 轴对齐- (2, 1) 的 x 轴应与 (2, 2) 的 x 轴对齐

Answer 1

我尝试了 renato vitolo 接受的答案，但对齐在我的机器上不起作用。但我随后发现了一个更简单的解决方案:egg 包(在 CRAN 上可用)。 egg 提供了一个名为 ggarrange 的 grid.arrange 版本，它采用类似的参数但可以很好地对齐轴。在 OP 的代码中，我只需添加 library(egg)、library(dplyr)，然后将 grid.arrange 替换为 ggarrange(已使用 install.packages("egg") 安装了 egg)。

完整代码:

library(gridExtra)
library(cowplot)
library(egg)
library(dplyr)

# generate some data
df_hm = cbind(
  expand.grid(
    rows = sample(letters, 10), 
    cols = sample(LETTERS, 10)
  ), 
  value = rnorm(100)
)

# plot the heatmap
gg_hm = df_hm %>% 
  ggplot(aes(x = rows, y = cols, fill = value)) + 
  geom_tile() + 
  theme(legend.position = "bottom")

gg_rows = df_hm %>% 
  group_by(rows) %>% 
  summarize(value = mean(value)) %>% 
  ggplot(aes(x = rows,y = value)) + 
  geom_bar(stat = "identity", position = "dodge") 

gg_cols = df_hm %>% 
  group_by(cols) %>% 
  summarize(value = mean(value)) %>% 
  ggplot(aes(x = cols, y = value))+ 
  geom_bar(stat = "identity", position = "dodge") + 
  coord_flip()

gg_empty = df_hm %>% 
  ggplot(aes(x = cols, y = value)) +
  geom_blank() +
  theme(axis.text = element_blank(),
        axis.title = element_blank(),
        line = element_blank(),
        panel.background = element_blank())

ggarrange(
  gg_rows, gg_empty, gg_hm, gg_cols,
  nrow = 2, ncol = 2, widths = c(3, 1), heights = c(1, 3)
)

输出: