Cambiar el orden de un diagrama de barras apiladas con ggplot2

Problema

Deseamos controlar el orden en el que se apilan las barras de un diagrama de barras en ggplot2.

• Gráfico original
Por defecto las barras se apilan alfabéticamente en función de la variable usada en el argumento fill.

library(reshape2)
library(ggplot2)
ra.melt <- melt(ra)
p <- ggplot(ra.melt, aes(x = variable, y = value))
p + geom_bar(aes(fill = quality), stat = "identity") + 
  labs(x = "group", y = "percentage (%)")

Comprobamos los niveles de la variable quality.

levels(ra.melt$quality)

 [1] "A"     "B"     "C"     "D"     "E"     "F"    
 [7] "G"     "H"     "I"     "J"     "K"     "L"    
[13] "M"     "other"

• Datos originales

ra <- structure(list(quality = structure(c(2L, 6L, 13L, 1L, 7L, 5L, 
10L, 4L, 3L, 9L, 11L, 12L, 8L, 14L), .Label = c("A", "B", "C", 
"D", "E", "F", "G", "H", "I", "J", "K", "L", "M", "other"), class = "factor"), GY = c(25.5932389, 13.9567819, 11.7333, 8.6439, 7.8351008, 
    8.29, 6.059643, 3.8608517, 4.3277, 3.1710598, 1.2021144, 
    1.0986329, 0.9369271, 3.2907496), TH = c(23.0215577, 15.8975387, 
    10.5075, 7.8888, 8.8121504, 8.276, 6.4356378, 3.9147685, 
    5.4838, 3.9339386, 1.3921565, 0.742406, 0.8795013, 2.8142444
    ), SZ = c(21.217152, 16.0363831, 11.154, 8.5094, 7.945884, 
    10.2717, 5.4962929, 4.1381946, 5.1141, 3.4053352, 1.8472827, 
    1.0648934, 0.9792861, 2.820096), DZ = c(23.7548859, 14.4732136, 
    12.1292, 10.4722, 8.0838209, 7.0575, 5.144598, 4.174118, 
    5.4004, 3.1940577, 1.4427867, 1.1323864, 0.8986281, 2.6422047
    ), FP = c(19.9848602, 16.420735, 12.9683, 11.2059, 7.5143706, 
    6.8747, 5.8653262, 5.5412901, 4.182, 3.5347168, 1.3838249, 
    0.8811604, 0.8585596, 2.7842562)), .Names = c("quality", 
"GY", "TH", "SZ", "DZ", "FP"), class = "data.frame", row.names = c("1", 
"2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12", "13", 
"14"))

Solución

• En el orden original del data frame

ra.melt$quality <- factor(ra.melt$quality, levels = ra$quality)
p <- ggplot(ra.melt, aes(x = variable, y = value))
p + geom_bar(aes(fill = quality), stat = "identity") + 
    labs(x = "group", y = "percentage (%)")

factor(ra.melt$quality, levels = ra$quality)

 [1] B     F     M     A     G     E     J     D    
 [9] C     I     K     L     H     other B     F    
[17] M     A     G     E     J     D     C     I    
[25] K     L     H     other B     F     M     A    
[33] G     E     J     D     C     I     K     L    
[41] H     other B     F     M     A     G     E    
[49] J     D     C     I     K     L     H     other
[57] B     F     M     A     G     E     J     D    
[65] C     I     K     L     H     other
Levels: B F M A G E J D C I K L H other

• En el orden reverso del original

ra.melt$quality <- factor(ra.melt$quality, levels = rev(ra$quality))
p <- ggplot(ra.melt, aes(x = variable, y = value))
p + geom_bar(aes(fill = quality), stat = "identity") +  
    labs(x = "group", y = "percentage (%)")

factor(ra.melt$quality, levels = rev(ra$quality))

 [1] B     F     M     A     G     E     J     D    
 [9] C     I     K     L     H     other B     F    
[17] M     A     G     E     J     D     C     I    
[25] K     L     H     other B     F     M     A    
[33] G     E     J     D     C     I     K     L    
[41] H     other B     F     M     A     G     E    
[49] J     D     C     I     K     L     H     other
[57] B     F     M     A     G     E     J     D    
[65] C     I     K     L     H     other
Levels: other H L K I C D J E G A M F B

O alternativamente, usamos la función fct_rev

library(forcats)
p <- ggplot(ra.melt, aes(x = variable, y = value))
p + geom_bar(aes(fill = fct_rev(quality)), stat = "identity") + 
  labs(x = "group", y = "percentage (%)")

• Si deseamos invertir el orden de las barras pero no de la leyenda, usamos el position_stack(reverse = TRUE)

p <- ggplot(ra.melt, aes(x = variable, y = value))
p + geom_bar(aes(fill = fct_rev(quality)), stat = "identity", position = position_stack(reverse = TRUE)) + labs(x = "group", y = "percentage (%)")

Referencias

• geom_bar
• stacking bars
• stackoverflow

Entradas relacionadas

• Destacar una barra de un diagrama de barras con ggplot2
• Diagramas de caja con ggplot2 en R. Una introducción.
• Destacar subconjunto de puntos en un diagrama de dispersión con ggplot2
• Etiquetar los puntos de un diagrama de dispersión con ggplot2
• Etiquetar selectivamente los puntos de un diagrama de dispersión con ggplot2
• Diagramas de dispersión con el paquete car en R