在ggplot facet_grid中创建小计

Question

Is there are an elegant way to add subtotal facets on a facet_grid(var1~var2) plot applied to a grouped aggregation, where an aggregation is similar to a sum , mean , sd , etc.

The "total" facet should use the same aggregation as the other facets.

The plot below gives an example of using mean . There are 6 subtotal and 1 total facets all up. Each total facet provides the mean mpg at each carb value.

My approach below is clunky and verbose. An approach that is generalized to handle a general number of both x and y faceting/grouping variables is desirable. I could code up a function that does a whole lot of inelegant conditional calculations depending on the groupings in the aggregation, but I'm really looking for something convenient and simple. Tidyverse solutions prefered.

library(dplyr)
library(ggplot2)
# use mtcars data set
data <- mtcars

# aggregate data by grouping variables
aggregate_data<- data%>%
  group_by(gear,cyl, carb)%>%
  summarize(mpg=mean(mpg))%>%
  ungroup

# get total for gear
data_tot_cyl<- data%>%
  group_by(cyl, carb)%>%
  summarize(mpg=mean(mpg))%>%
  ungroup%>%
  mutate(gear='total')

# get total for cyl
data_tot_gear<- data%>%
  group_by(gear, carb)%>%
  summarize(mpg=mean(mpg))%>%
  ungroup%>%
  mutate(cyl='total')

# get total for total-total
data_tot_tot<- data%>%
  group_by(carb)%>%
  summarize(mpg=mean(mpg))%>%
  ungroup%>%
  mutate(cyl='total', gear='total')

# get data frame with all total's data.
new_data<-data_tot_tot%>%
  bind_rows(data_tot_gear%>%mutate(gear=as.character(gear)))%>%
  bind_rows(data_tot_cyl%>%mutate(cyl=as.character(cyl)))%>%
  bind_rows(aggregate_data%>%mutate_at(vars(gear, cyl), funs(as.character)))

# Arghh, gotta order the levels so total is at the end.
new_data$cyl <- factor(new_data$cyl, 
                         levels=c('4','6','8','total'),ordered=T)
new_data$gear <- factor(new_data$gear, 
                        levels=c('3','4','5','total'),ordered=T)

# Finally after over 20 additional lines of code, I get the 
# faceted plot with totals for x and y facets. 
p<-ggplot(new_data, aes(x=carb, y=mpg))+
  geom_bar(stat='identity')+
  facet_grid(cyl~gear)+
  geom_text(aes(label=round(mpg,1), y=0), 
            col='white', size=3,hjust=-0.3, angle=90)+
  ggtitle('Average MPG vs Num Carbs, by Num Cylinders & Num Gears')
print(p)

Answer 1

you want to use the margins option in your facet_grid() function. See the following:

p2 <- ggplot(aggregate_data, aes(x=1, y=value))+
  geom_point()+
  facet_grid(dist~scale, margins = TRUE)
p2

Answer 2

ggplot can do it directly from the original data frame. But the geom_text still displays the individual row values, rather than the aggregation.

p3<-ggplot(mtcars, aes(x=carb, y=mpg))+
  stat_summary(fun.y="mean", geom="bar")+
  facet_grid(cyl~gear, margins=T)+
  geom_text(aes(label=round(..y..,1)), 
            col='red', size=3,hjust=-0.3, angle=90)+
  ggtitle('Average MPG vs Num Carbs, by Num Cylinders & Num Gears')
print(p3)

And it allows you to group the aggregation by any facet_grid combination. Eg

p4<-ggplot(mtcars, aes(x=carb, y=mpg))+
  stat_summary(fun.y="mean", geom="bar")+
  facet_grid(am+vs~gear, margins=T)+
  geom_text(aes(label=round(..y..,1)), 
            col='red', size=3,hjust=-0.3, angle=90)+
  ggtitle('Average MPG vs Num Carbs, by Num Gears and vs & am')
print(p4)