ggplot热图格线格式化：geom_tile和geom_rect

Question

ggplot热图格线格式化：geom_tile和geom_rect

6

我一直在创建热力图，但几天来一直无法得到网格线的最终格式。请参见下面的代码和附加的图表。我正在尝试使用geom_tile()对齐热力图块的网格线，以便每个块填充网格的内部以盒状方式。我能够使用geom_raster()对齐网格线，但是y轴标签刻度要么位于瓷砖的顶部，要么位于底部，而我需要它在中心处打勾(见红色高亮)，此外，我无法让geom_raster在瓷砖周围包裹白色线条边框，因此原始数据集中的色块看起来有些杂乱。非常感谢任何关于格式化代码的帮助。谢谢！

#The data set in long format 


y<- c("A","A","A","A","B","B","B","B","B","C","C","C","D","D","D")
    x<- c("2020-03-01","2020-03-15","2020-03-18","2020-03-18","2020-03-01","2020-03-01","2020-03-01","2020-03-01","2020-03-05","2020-03-06","2020-03-05","2020-03-05","2020-03-20","2020-03-20","2020-03-21")
    v<-data.frame(y,x)

#approach 1 using geom_tile but gridline does not align with borders of the tiles 
    v%>%
      count(y,x,drop=FALSE)%>%
      arrange(n)%>%
      ggplot(aes(x=x,y=fct_reorder(y,n,sum)))+
      geom_tile(aes(fill=n),color="white", size=0.25)

我尝试过运行来自另一篇帖子类似的代码，但无法正确运行。我认为这是因为我的x变量是y变量的计数变量，因此无法格式化为因子变量以指定geom_rect（）中的xmin和xmax。

#approach 2 using geom_raster but y-axis label can't tick at the center of tiles and there's no border around the tile to differentiate between tiles. 

v%>%
  count(y,x,drop=FALSE)%>%
  arrange(n)%>%
  ggplot()+
  geom_raster(aes(x=x,y=fct_reorder(y,n,sum),fill=n),hjust=0,vjust=0)

- DHR

2个回答

2

这有点取巧。我的方法是将分类变量转换为数字，这会在绘图中添加与瓷砖对齐的次要网格线。为了去掉主要的网格线，我只需使用 theme()。缺点是必须手动设置断点和标签。

library(ggplot2)
library(dplyr)
library(forcats)

v1 <- v %>%
  count(y,x,drop=FALSE)%>%
  arrange(n) %>%
  mutate(y = fct_reorder(y, n, sum),
         y1 = as.integer(y),
         x = factor(x),
         x1 = as.integer(x))

labels_y <- levels(v1$y)
breaks_y <- seq_along(labels_y)

labels_x <- levels(v1$x)
breaks_x <- seq_along(labels_x)

ggplot(v1, aes(x=x1, y=y1))+
  geom_tile(aes(fill=n), color="white", size=0.25) + 
  scale_y_continuous(breaks = breaks_y, labels = labels_y) +
  scale_x_continuous(breaks = breaks_x, labels = labels_x) +
  theme(panel.grid.major = element_blank())

^{由reprex包 (v0.3.0)于2020年5月23日创建}

编辑：检查了长变量名称

y<- c("John Doe","John Doe","John Doe","John Doe","Mary Jane","Mary Jane","Mary Jane","Mary Jane","Mary Jane","C","C","C","D","D","D")
x<- c("2020-03-01","2020-03-15","2020-03-18","2020-03-18","2020-03-01","2020-03-01","2020-03-01","2020-03-01","2020-03-05","2020-03-06","2020-03-05","2020-03-05","2020-03-20","2020-03-20","2020-03-21")
v<-data.frame(y,x)

^{创建于2020年5月23日，使用reprex软件包（v0.3.0）}

- stefan

谢谢，Stefan。然而，在实际数据集中，我的y变量具有更长的字符标签（例如“Mary Jane”，“John Doe”等），因此无法转换为整数，并显示为y1中的NA。在这种情况下，您有什么建议？对不起，我应该考虑到这一点并创建样本数据集。 - DHR

嗨@DHR。你在将数据转换为数值之前是否已将其转换为因子？看一下我的编辑内容。我将A改成了John Doe，B改成了Mary Jane。我的代码仍适用于这种情况。 - stefan

网页内容由stack overflow 提供, 点击上面的

可以查看英文原文，
原文链接

- PRZ · Accepted Answer

我认为保留刻度和网格线的位置是有意义的。要实现你想要的效果，我建议扩展数据包含所有可能的组合，并将na.value设置为中性填充颜色：

# all possible combinations
all <- v %>% expand(y, x)

# join with all, n will be NA for obs. in all that are not present in v
v = v %>% group_by_at(vars(y, x)) %>% 
    summarize(n = n()) %>% right_join(all)

ggplot(data = v, 
       aes(x=x, y=fct_reorder(y,n, function(x) sum(x, na.rm = T))))+ # note that you must account for the NA values now 
geom_tile(aes(fill=n), color="white",
        size=0.25) +
scale_fill_continuous(na.value = 'grey90') +
scale_x_discrete(expand = c(0,0)) +
scale_y_discrete(expand = c(0,0))