如何在R中使用facet_wrap并排绘制箱线图？

Question

I was looking for a solution to draw side by side boxplot using facet_wrap in R . Though there are lots of good solutions, however, I didn't come across any that i wanted. I decided to draw a picture of the plot that i would like to see of my two data.frame . Data.frame C has my calibration data for the four models of different meterics (ie, KGE, NSE, PBIAS, and R-Sq) while Data.frame V has my validation data. I want to see a separate plot of each metrics using facet_wrap of the ggplot2 functionality. Below is what i have done so far but its not taking me even closer.

graphics.off()
rm(list = ls())

library(tidyverse)

C = data.frame(KGE_M1 = runif(3, 0, 0.5), NSE_M1 = runif(3,0,0.5), R_Sq_M1 = runif(3,-1,0.3), PBIAS_M1 = runif(3, -0.25, 0.25),
               KGE_M2 = runif(3, 0.2, 0.7), NSE_M2 = runif(3,0.2,0.7), R_Sq_M2 = runif(3,-0.5,0.7), PBIAS_M2 = runif(3, -0.15, 0.15),
               KGE_M3 = runif(3, 0.3, 0.8), NSE_M3 = runif(3,0.3,0.8), R_Sq_M3 = runif(3,0.3,0.8), PBIAS_M3 = runif(3, -0.10, 0.10),
               KGE_M4 = runif(3, 0.5, 1), NSE_M4 = runif(3,0.5,1), R_Sq_M4 = runif(3,0.5,1), PBIAS_M4 = runif(3, -0.05, 0.05),
               Cal = rep("Calibration", 3))

V = data.frame(KGE_M1 = runif(3, 0, 0.5), NSE_M1 = runif(3,0,0.5), R_Sq_M1 = runif(3,-1,0.3), PBIAS_M1 = runif(3, -0.25, 0.25),
               KGE_M2 = runif(3, 0.2, 0.7), NSE_M2 = runif(3,0.2,0.7), R_Sq_M2 = runif(3,-0.5,0.7), PBIAS_M2 = runif(3, -0.15, 0.15),
               KGE_M3 = runif(3, 0.3, 0.8), NSE_M3 = runif(3,0.3,0.8), R_Sq_M3 = runif(3,0.3,0.8), PBIAS_M3 = runif(3, -0.10, 0.10),
               KGE_M4 = runif(3, 0.5, 1), NSE_M4 = runif(3,0.5,1), R_Sq_M4 = runif(3,0.5,1), PBIAS_M4 = runif(3, -0.05, 0.05),
               Val = rep("Validation", 3))

C = gather(C, key = "Variable", value = "Value", -Cal)
V = gather(V, key = "Variable", value = "Value", -Val)

ggplot(data = C)+
  geom_boxplot(aes(x= Variable, y = Value))
  + facet_wrap(~Variable)

I would like to see a plot like below

Answer 1

So here's a way with which you can do the required;

First we create the data that you have;

library(tidyverse)

# Creating first dataframe
C <- 
  data.frame(
    KGE_M1 = runif(3, 0, 0.5), 
    NSE_M1 = runif(3,0,0.5), 
    R_Sq_M1 = runif(3,-1,0.3), 
    PBIAS_M1 = runif(3, -0.25, 0.25),
    KGE_M2 = runif(3, 0.2, 0.7), 
    NSE_M2 = runif(3,0.2,0.7), 
    R_Sq_M2 = runif(3,-0.5,0.7), 
    PBIAS_M2 = runif(3, -0.15, 0.15),
    KGE_M3 = runif(3, 0.3, 0.8), 
    NSE_M3 = runif(3,0.3,0.8), 
    R_Sq_M3 = runif(3,0.3,0.8), 
    PBIAS_M3 = runif(3, -0.10, 0.10),
    KGE_M4 = runif(3, 0.5, 1), 
    NSE_M4 = runif(3,0.5,1), 
    R_Sq_M4 = runif(3,0.5,1), 
    PBIAS_M4 = runif(3, -0.05, 0.05),
    Cal = rep("Calibration", 3),
    stringsAsFactors = FALSE)

# Creating second dataframe
V <- 
  data.frame(
    KGE_M1 = runif(3, 0, 0.5), 
    NSE_M1 = runif(3,0,0.5), 
    R_Sq_M1 = runif(3,-1,0.3), 
    PBIAS_M1 = runif(3, -0.25, 0.25),
    KGE_M2 = runif(3, 0.2, 0.7), 
    NSE_M2 = runif(3,0.2,0.7), 
    R_Sq_M2 = runif(3,-0.5,0.7), 
    PBIAS_M2 = runif(3, -0.15, 0.15),
    KGE_M3 = runif(3, 0.3, 0.8), 
    NSE_M3 = runif(3,0.3,0.8), 
    R_Sq_M3 = runif(3,0.3,0.8), 
    PBIAS_M3 = runif(3, -0.10, 0.10),
    KGE_M4 = runif(3, 0.5, 1), 
    NSE_M4 = runif(3,0.5,1), 
    R_Sq_M4 = runif(3,0.5,1), 
    PBIAS_M4 = runif(3, -0.05, 0.05),
    Val = rep("Validation", 3),
    stringsAsFactors = FALSE)

Now we change format of the data and visualize it;

# Rename the variable to make it same
C <- rename(C, Identifier = Cal)
V <- rename(V, Identifier = Val)

data <- 
  # First we bind the two datasets
  bind_rows(C, V) %>%
  # We convert from wide format to long format
  gather(key = "Variable", value = "Value", -Identifier) %>%
  # We separate Variable into 2 columns at the last underscore
  separate(Variable, into = c("Variable", "Number"), sep = "_(?=[^_]+$)")

data %>%
  ggplot()+
  geom_boxplot(aes(x = Number, y = Value, 
                   group  = interaction(Identifier, Number), fill = Identifier)) + 
  facet_wrap(~Variable)

Answer 2

I think you need to split your Variable before plotting in order to have one variable for M1, M2, M3 M4 and one variable for your conditions:

library(tidyverse)
C2 <- C %>% pivot_longer(., -Cal, names_to = "Variable", values_to = "Value") %>%
  group_by(Variable) %>%
  mutate(Variable2 = unlist(strsplit(Variable, "_M"))[2]) %>%
  mutate(Variable2 = paste0("Cal_M",Variable2)) %>%
  mutate(Variable1 = unlist(strsplit(Variable,"_M"))[1])  %>%
  rename(., Type = Cal)

# A tibble: 6 x 5
# Groups:   Variable [6]
  Type        Variable  Value Variable2 Variable1
  <fct>       <chr>     <dbl> <chr>     <chr>    
1 Calibration KGE_M1    0.246 Cal_M1    KGE      
2 Calibration NSE_M1    0.476 Cal_M1    NSE      
3 Calibration R_Sq_M1  -0.978 Cal_M1    R_Sq     
4 Calibration PBIAS_M1  0.117 Cal_M1    PBIAS    
5 Calibration KGE_M2    0.544 Cal_M2    KGE      
6 Calibration NSE_M2    0.270 Cal_M2    NSE   

Now, we are doing the same thing for the dataset V

V2 <- V %>% pivot_longer(., -Val, names_to = "Variable", values_to = "Value") %>%
  group_by(Variable) %>%
  mutate(Variable2 = unlist(strsplit(Variable, "_M"))[2]) %>%
  mutate(Variable2 = paste0("Val_M",Variable2)) %>%
  mutate(Variable1 = unlist(strsplit(Variable,"_M"))[1]) %>%
  rename(., Type = Val)

# A tibble: 6 x 5
# Groups:   Variable [6]
  Type       Variable   Value Variable2 Variable1
  <fct>      <chr>      <dbl> <chr>     <chr>    
1 Validation KGE_M1    0.459  Val_M1    KGE      
2 Validation NSE_M1    0.105  Val_M1    NSE      
3 Validation R_Sq_M1  -0.435  Val_M1    R_Sq     
4 Validation PBIAS_M1  0.0281 Val_M1    PBIAS    
5 Validation KGE_M2    0.625  Val_M2    KGE      
6 Validation NSE_M2    0.332  Val_M2    NSE    

We can now bind them together:

DF <- rbind(C2,V2)

Then, we can plot:

ggplot(DF, aes(x = Variable2, y = Value))+
  geom_boxplot()+
  facet_wrap(.~Variable1, scales = "free")+
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

EDIT: Rename x axis, add empty column to separate Calibration and Validation values

To add an empty space between Calibration and Validation, you can simply add empty rows for each conditions of the Variable1 like this:

DF <- as.data.frame(DF) %>% add_row(Type = rep("Empty",4),
                     Variable = rep("Empty",4),
                     Value = rep(NA,4),
                     Variable2 = rep("Empty",4),
                     Variable1 = unique(DF$Variable1))

Also, if you want to rename the x axis labels, you can use scale_x_discrete

ggplot(DF, aes(x = Variable2, y = Value, fill = Type))+
  geom_boxplot()+
  facet_wrap(.~Variable1, scales = "free")+
  theme(axis.text.x = element_text(angle = 45, hjust = 1))+
  scale_x_discrete(labels = c("M1","M2","M3","M4","","M1","M2","M3","M4"))

Does it look what you are expecting ?