Scaling multiple pie charts in a grid according to their size in R using ggplot2

Question

I want to create a grid containing 9 pie charts (3x3) with each chart being scaled according to its size. Using ggplot2 and cowplot I was able to create what I was looking for, but I could not do the scaling. Am I just overlooking a function or should I use another package? I also tried grid.arrange from the gridExtra package and ggplot's facet_grid function, but both did not produce what I am looking for.

I also found a similar question (Pie charts in ggplot2 with variable pie sizes) that used facet_grid. Unfortunately, this does not work in my case since I am not comparing two variables in respect of all possible outcomes.

So this is my sample code:

#sample data
x <- data.frame(c("group01", "group01", "group02", "group02", "group03", "group03",
                  "group04", "group04", "group05", "group05", "group06", "group06",
                  "group07", "group07", "group08", "group08", "group09", "group09"),
                c("w","m"),
                c(8,8,6,10,26,19,27,85,113,70,161,159,127,197,179,170,1042,1230),
                c(1,1,1,1,3,3,7,7,11,11,20,20,20,20,22,22,142,142))
colnames(x) <- c("group", "sex", "data", "scale")
#I have divided the group size by the smallest group (group01, 16 people) in order to receive the scaling-variable.
#Please note that I doubled the values here for simplicity-reasons for both men and women per group (for plot-scaling only one value is needed that I calculate 
#seperately in the original data in the plot-scaling part underneath).
#In this example I am also going to use the scaling-variable as indicator of the sequence of the plots.

library(ggplot2)
library(cowplot)

#Then I create 9 pie-charts, each one containing one group and showing the quantity of men vs. women in a very simplistic style 
#(only the name of the group showing; color of each sex is explained seperately in the according text)
p1 <- ggplot(x[c(1,2),], aes("", y = data, fill = factor(sex), x$scale[1]))+
  geom_bar(width = 4, stat="identity") + coord_polar("y", start = 0, direction = 1)+
  ggtitle(label=x$group[1])+
  theme_classic()+theme(legend.position = "none")+
  theme(axis.title=element_blank(),axis.line=element_blank(),axis.ticks=element_blank(),axis.text=element_blank(),plot.background = element_blank(),
        plot.title=element_text(color="black",size=10,face="plain",hjust=0.5))
p2 <- ggplot(x[c(3,4),], aes("", y = data, fill = factor(sex), x$scale[3]))+
  geom_bar(width = 4, stat="identity") + coord_polar("y", start = 0, direction = 1)+
  ggtitle(label=x$group[3])+
  theme_classic()+theme(legend.position = "none")+
  theme(axis.title=element_blank(),axis.line=element_blank(),axis.ticks=element_blank(),axis.text=element_blank(),plot.background = element_blank(),
        plot.title=element_text(color="black",size=10,face="plain",hjust=0.5))
p3 <- ggplot(x[c(5,6),], aes("", y = data, fill = factor(sex), x$scale[5]))+
  geom_bar(width = 4, stat="identity") + coord_polar("y", start = 0, direction = 1)+
  ggtitle(label=x$group[5])+
  theme_classic()+theme(legend.position = "none")+
  theme(axis.title=element_blank(),axis.line=element_blank(),axis.ticks=element_blank(),axis.text=element_blank(),plot.background = element_blank(),
        plot.title=element_text(color="black",size=10,face="plain",hjust=0.5))
p4 <- ggplot(x[c(7,8),], aes("", y = data, fill = factor(sex), x$scale[7]))+
  geom_bar(width = 4, stat="identity") + coord_polar("y", start = 0, direction = 1)+
  ggtitle(label=x$group[7])+
  theme_classic()+theme(legend.position = "none")+
  theme(axis.title=element_blank(),axis.line=element_blank(),axis.ticks=element_blank(),axis.text=element_blank(),plot.background = element_blank(),
        plot.title=element_text(color="black",size=10,face="plain",hjust=0.5))
p5 <- ggplot(x[c(9,10),], aes("", y = data, fill = factor(sex), x$scale[9]))+
  geom_bar(width = 4, stat="identity") + coord_polar("y", start = 0, direction = 1)+
  ggtitle(label=x$group[9])+
  theme_classic()+theme(legend.position = "none")+
  theme(axis.title=element_blank(),axis.line=element_blank(),axis.ticks=element_blank(),axis.text=element_blank(),plot.background = element_blank(),
        plot.title=element_text(color="black",size=10,face="plain",hjust=0.5))
p6 <- ggplot(x[c(11,12),], aes("", y = data, fill = factor(sex), x$scale[11]))+
  geom_bar(width = 4, stat="identity") + coord_polar("y", start = 0, direction = 1)+
  ggtitle(label=x$group[11])+
  theme_classic()+theme(legend.position = "none")+
  theme(axis.title=element_blank(),axis.line=element_blank(),axis.ticks=element_blank(),axis.text=element_blank(),plot.background = element_blank(),
        plot.title=element_text(color="black",size=10,face="plain",hjust=0.5))
p7 <- ggplot(x[c(13,14),], aes("", y = data, fill = factor(sex), x$scale[13]))+
  geom_bar(width = 4, stat="identity") + coord_polar("y", start = 0, direction = 1)+
  ggtitle(label=x$group[13])+
  theme_classic()+theme(legend.position = "none")+
  theme(axis.title=element_blank(),axis.line=element_blank(),axis.ticks=element_blank(),axis.text=element_blank(),plot.background = element_blank(),
        plot.title=element_text(color="black",size=10,face="plain",hjust=0.5))
p8 <- ggplot(x[c(15,16),], aes("", y = data, fill = factor(sex), x$scale[15]))+
  geom_bar(width = 4, stat="identity") + coord_polar("y", start = 0, direction = 1)+
  ggtitle(label=x$group[15])+
  theme_classic()+theme(legend.position = "none")+
  theme(axis.title=element_blank(),axis.line=element_blank(),axis.ticks=element_blank(),axis.text=element_blank(),plot.background = element_blank(),
        plot.title=element_text(color="black",size=10,face="plain",hjust=0.5))
p9 <- ggplot(x[c(17,18),], aes("", y = data, fill = factor(sex), x$scale[17]))+
  geom_bar(width = 4, stat="identity") + coord_polar("y", start = 0, direction = 1)+
  ggtitle(label=x$group[17])+
  theme_classic()+theme(legend.position = "none")+
  theme(axis.title=element_blank(),axis.line=element_blank(),axis.ticks=element_blank(),axis.text=element_blank(),plot.background = element_blank(),
        plot.title=element_text(color="black",size=10,face="plain",hjust=0.5))

#Using cowplot, I create a grid that contains my plots
plot_grid(p1,p2,p3,p4,p5,p6,p7,p8,p9, align = "h", ncol = 3, nrow = 3)

#But now I want to scale the size of the plots according to their real group size (e.g.
#group01 with 16 people vs. group09 with more than 2000 people)


#In this context, ggplot's facet_grid function produces similar results of what I want to get, 
#but since it looks at the data as a whole instead of separating groups from each other, it does not show 
#complete pie charts per group

#So is there a possibility to scale each of the 9 charts according to their group size?

This is what plot_grid produces: pie-charts without scaling

Using the rel_widths argument I could only adjust the scaling, but was not able to maintain the 3x3 grid.

plot_grid(p1,p2,p3,p4,p5,p6,p7,p8,p9, 
          align="h",ncol=(nrow(x)/2),          
          rel_widths = c(x$scale[1], 
                         x$scale[3], 
                         x$scale[5], 
                         x$scale[7], 
                         x$scale[9], 
                         x$scale[11], 
                         x$scale[13], 
                         x$scale[15], 
                         x$scale[17]))

This is what adjusting rel_widths does:

pie-charts with scaling, but without grid

In conclusion, what I need is a mixture of both: scaled pie-charts in a grid.

J. Ring · Answer 1 · 2017-10-11T12:21:08.840

You are on the right path. Problem is that you are assigning c(...) values for 3x3 output. Here are two ways you could do:

# Option 1    
# individually for each row
        plot_upper <- plot_grid(p1, p2, p3, labels = "", ncol = 3, rel_widths = c(1, 1.1, 1.2))
        plot_middle <- plot_grid(p4, p5, p6, labels = "", ncol = 3, rel_widths = c(1.3, .3, 1.3))
        plot_lower <- plot_grid(p7, p8, p9, labels = "", ncol = 3, rel_widths = c(1.2, 1.1, 1))
        plot_grid(plot_upper, plot_middle, plot_lower, ncol = 1, rel_heights = c(1, 2.5, 1.7))
# Option 2
# Set size matrix
    sizes <- matrix(c(x$scale[1], 
              x$scale[3], 
              x$scale[5], 
              x$scale[7], 
              x$scale[9], 
              x$scale[11], 
              x$scale[13], 
              x$scale[15], 
              x$scale[17]), ncol = 3)
    plot_grid(p1,p2,p3,p4,p5,p6,p7,p8,p9, align = "h", ncol = 3, nrow = 3, rel_widths = sizes, rel_heights = sizes)

Here is a link for documentation and here are some examples.

So if you adjust your scales to matrix and insert it in option 2 you should acquire what you wish. Also, ...cbind(... when defining your data.frame is unnecessary.

Thanks! This produces a grid that is almost what I am looking for. Unfortunately, it also creates overlap between the titles of the charts when there is no enough space in between. I deleted cbind(...) from the code. Thanks! — alex_555, Oct 11 '17 at 09:03
Had mistakenly defined the matrix `sizes` before. Corrected it for your case, now should work as intended. — J. Ring, Oct 11 '17 at 12:23

Marco Sandri · Accepted Answer · 2017-10-11T13:40:27.280

1

What about this ?

x$scale <- as.numeric(x$scale)
x$data <- as.numeric(x$data)
x$group <- factor(x$group, levels=levels(x$group)[order(x$scale[seq(1,nrow(x),2)])])

ggplot(x, aes(x=scale/2, y = data, fill = factor(sex), width=scale))+
  geom_bar(position="fill", stat="identity") + coord_polar("y")+
  facet_wrap( ~ group, nrow=3) +
  theme_classic()+theme(legend.position = "none")+
  theme(axis.title=element_blank(), axis.line=element_blank(),
        axis.ticks=element_blank(), axis.text=element_blank(),
        plot.background = element_blank(), 
        plot.title=element_text(color="black",size=10,face="plain",hjust=0.5),
        strip.background = element_blank(), 
        strip.text.x = element_text(color = "transparent") )

edited Oct 11 '17 at 13:40

answered Oct 11 '17 at 08:49

Marco Sandri

23,289
7
54
58

Great! This is almost what I am looking for! Unfortunately, it throws off the order. Is there a possibility to order them by group size? – alex_555 Oct 11 '17 at 09:02
This works fine with my sample data, but it does not want to do something as simple as creating a factor out of my original group names. Weird. I am quite sure that I am just overlooking a little typing error or something similar. So, I will solve the problem soon...hopefully. Thank you! – alex_555 Oct 11 '17 at 09:45
Done. Thanks for your help again! – alex_555 Oct 11 '17 at 10:21

Scaling multiple pie charts in a grid according to their size in R using ggplot2

2 Answers2