设置＆＃34;空间＆＃34;在facet_wrap中像facet_grid一样

Question

我需要不同宽度的方面;左图显示实验的动态范围，右图显示测试条件。有没有办法让facet_wrap有自由x和y标度？它可以在facet_grid中使用，但即使使用scale =＆＃34; free＆＃34;，也有一个固定的y标度。 facet_wrap允许自由y比例，但x比例似乎是固定的。几年前在谷歌页面上发布了同样的问题，但答案并不令人满意。 https://groups.google.com/forum/#!topic/ggplot2/1RwkCcTRBAw

对不起，如果这里也重复一遍;任何帮助都会非常感激！

mdf <- read.table(text="
   strain val     type
1       1 0.0000  sample
2       1 0.0140  sample
3       1 0.0175  sample
4       2 0.0025  sample
5       2 0.0260  sample
6       2 0.0105  sample
7       3 0.0190  sample
8       3 0.0725  sample
9       3 0.0390  sample
10      4 0.0560  sample
11      4 0.0695  sample
12      4 0.0605  sample
13      5 0.0735  sample
14      5 0.1065  sample
15      5 0.0890  sample
16      6 0.1135  sample
17      6 0.2105  sample
18      6 0.1410  sample
19      7 0.1360  sample
20      7 0.2610  sample
21      7 0.1740  sample
22      8 0.3850 control
23      8 0.7580 control
24      8 0.5230 control
25      9 0.5230 control
26      9 0.5860 control
27      9 0.7240 control")

library(ggplot2)

p<-ggplot(mdf, aes(reorder(strain, val), val))+
  labs(x="Strain", y="intensity")+
  geom_boxplot()+
  geom_point()+
  facet_grid(~type, scales ="free", space="free_x")
p
##  free x, fixed y.  why?

q<-ggplot(mdf, aes(reorder(strain, val), val))+
  labs(x="Strain", y="intensity")+
  geom_boxplot()+
  geom_point()+
  facet_wrap(~type, scales ="free")
q
##  free y, fixed x.  why?

Answer 1

我无法确定，但我认为答案是否定的 - 使用ggplot2命令。我认为这不是一个好主意，因为读者可能并不明白y轴上的刻度是不同的。但是，如果必须有绘图，则可以使用ggplot grob布局调整q绘图的面板宽度。请注意，第一个面板有两个x值，第二个面板有七个x值。因此，将面板的默认宽度分别更改为2null和7null。

编辑：更新到ggplot2 2.2.0

library(ggplot2)
library(grid)
# get mdf data frame from the question

# Your q plot
q <- ggplot(mdf, aes(factor(strain), val)) +
  labs(x = "Strain", y = "intensity") +
  geom_boxplot() +
  geom_point() +
  facet_wrap( ~ type, scales = "free")
q

# Get the ggplot grob
gt = ggplotGrob(q)

# Check for the widths - you need to change the two that are set to 1null
gt$widths
# The required widths are 4 and 8

# Replace the default widths with relative widths:
gt$widths[4] = unit(2, "null")
gt$widths[8] = unit(7, "null")

# Draw the plot
grid.newpage()
grid.draw(gt)

# I think it is better to have some extra space between the two panels
gt$widths[5] = unit(1, "cm")
grid.newpage()
grid.draw(gt)

或者，让R确定相对宽度和面板。

gt = ggplotGrob(q)

# From 'dfm', get the number of 'strain' for each 'type'.
# That is, the number x-breaks in each panel.
library(dplyr)
N <- mdf %>% group_by(type) %>% 
     summarise(count = length(unique(strain))) %>% 
     `[[`(2)

# Get the column index in the gt layout corresponding to the panels.
panelI <- gt$layout$l[grepl("panel", gt$layout$name)]

# Replace the default panel widths with relative heights.
gt$widths[panelI] <- unit(N, "null")

# Add extra width between panels (assuming two panels)
gt$widths[panelI[1] + 1] = unit(1, "cm")

## Draw gt
grid.newpage()
grid.draw(gt)

Answer 2

此外，对于那些试图使用@ Sandy的答案的人来说：dplyr：

library(dplyr)
N<-mdf%>% group_by(type)%>% summarise(count = length(unique(strain)))

# Get the column index in the gt layout corresponding to the panels.
panelI <- gt$layout$l[grepl("panel", gt$layout$name)]

# Replace the default panel widths with relative heights.
gt$widths[panelI] <- lapply(N$count, unit, "null")