How to control space between stack bars in ggplot2

问题:

I'm plotting a stack barplot in ggplot2. My dataset is like,

var1         var2   var3       value
treatment1  group_1 C8.0    0.010056478
treatment2  group_1 C8.0    0.009382918
treatment3  group_2 C8.0    0.003014983
treatment4  group_2 C8.0    0.005349631
treatment5  group_2 C8.0    0.005349631

var1 contains 5 treatments, these five treatments belong to two groups in var2, and each treatment has 14 measurements in var3, their value stored in value.

I want to make a plot to compare these five treatments, and their measurements. so I plot with stack bar plot like this figure:

My code:

library(ggplot2)
colourCount = length(unique(mydata$var3))
getPalette = colorRampPalette(brewer.pal(14, "YlGnBu")) #get more color from palette

ggplot(data=mydata, aes(x=var1, y=value, fill=var3))+
  geom_bar(stat="identity", position="stack", colour="black", width=.2)+
  *#geom_errorbar(aes(ymax=var3+se, ymin=var3-se, width=.1))+*
  scale_fill_manual(values = getPalette(colourCount))+

  scale_y_continuous(expand = c(0, 0))+
  mytheme

How could I group the first two stacked columns together, and the other three columns together? Because they belong to two groups in var2.

回答1:

The "duplicate question" comments above will lead you to an answer like this one:

library(dplyr)
library(ggplot2)
dummydf <- expand.grid(var1 = paste0("trt", 1:5),
            var3 = paste0("C_", 11:15)) %>% 
  mutate(value = runif(length(var1)),
         var2 = ifelse(var1 %in% c("trt1", "trt2"), "grp1", "grp2"))


  ggplot(dummydf, aes(var1, value, fill = var3)) +
  geom_col(position = "stack") +
  facet_grid(~var2, scales = "free_x", space = "free_x") 

And this solution is sometimes great! The advantages are:

1. it's simple to implement
2. contains the labels for the hierarchical grouping at the top
3. generally looks nice
4. is easily customizable.

For example:

  ggplot(dummydf, aes(var1, value, fill = var3)) +
  geom_col(position = "stack") +
  facet_grid(~var2, scales = "free_x", space = "free_x") +
  theme(panel.spacing = unit(3, "cm"),
        strip.text = element_text(size = 12, family = "mono"))

The main disadvantages to this method:

1. If this is already part of a facet panelled plot, it makes the whole thing cluttered.
2. If your hierarchical groups are obvious from the treatments, you may not need explicit labelling, just a quick visual distinction. For example, say the groups are control/intervention, and your treatments are "no drug, placebo" and "drugs 1, 2, and 3".

So here's an alternate method:

dummydf %>% 
  bind_rows(data_frame(var1 = "trt99")) %>% 
  ggplot(aes(var1, value, fill = var3)) +
  geom_col(position = "stack") +
  scale_x_discrete(limits = c("trt1", "trt2", "trt99", "trt3", "trt4", "trt5"),
                   breaks = c("trt1", "trt2",      NA, "trt3", "trt4", "trt5"),
                   labels = c("trt1", "trt2",      "", "trt3", "trt4", "trt5"))

This solution has its own drawbacks, primarily that you can only customize the space in a limited way. You can create a "false" bar equal to an integer multiple of the widths of the bars you've already got by adding additional false levels to your limits, breaks, and labels. But you can't create a space that's only half a bar wide.

You could provide additional information in the false bar space though:

1. Adding a text annotation in the plot area
2. Replacing the NA and "" in breaks and labels with trt99 and "<-group1 | group2->" or something similar.

回答2:

I think you just need to create a new column to your data set which labels everything is "treatment3" and "not treatment3". I used dplyr package:

df1 = dplyr::mutate(mydata, 
    var4 = ifelse(var1 == "treatment3", "treatment3", "not treatmeant3"))