2025-03-13



Introduction

ggpubr package provides some easy-to-use functions for creating and customizing ggplot2 based plots.

In particular, it contains:

  • already designed plots that you can explore @ https://rpkgs.datanovia.com/ggpubr/, including the table plot

  • several statistical methods for adding correlation coefficients, p-value evaluating the difference of distributions and so on…

  • additional methods for gglot2 objects, like arranging more than one plot in the same panel



Correlation

If we have two continous variables and we want to see the relation between them, we can make a scatter plot. Let’s say that by plotting we notice a possible correlation between the two. We want to see if the correlation we see is statistically significant directly on the plot. In this case we can use stat_cor(). We will use, once again, the mpg and we want to see the relation between the miles per gallon on highway and in the city:

library(ggpubr)
library(ggplot2)
ggplot(mpg, aes(x=hwy, y=cty))+geom_point()+theme_bw()+stat_cor()

As we can see on the top of the graph are indicated the correlation coefficient, R and the statistic significance, the p-value p.



Comparing the means of two independent groups

If we have a continous variable and we want to compare the values it assumes in different groups, we can make a boxplot. Then we may want to evaluate if there is a statistical difference between values assumed in each group. For example, we may want to know if there is a significant difference between the miles per gallon in the city between cars produced in 1999 and 2008 We will add stat_compare_means() to evaluate the difference. Notice: as the year is a continous variable in the data frame, we will first transform it in a discrete one by changing its type in character. We will use notches in boxplot to better see differences.

ggplot(mpg, aes(x=as.character(year), y=cty))+geom_boxplot(notch = TRUE,width=0.6)+theme_bw()+stat_compare_means(label.x = 1.5)

This means that there is no statistical difference (p-value=0.92).

If we compare the city miles per gallon for the suv and compact classes, we obtain a low p-value, indicating that the two groups are statistically different.

ggplot(subset(mpg,class%in%c("suv","compact")),aes(x=class,y=cty))+
  geom_boxplot(notch = TRUE,width=0.6)+theme_bw()+
  stat_compare_means(label.x = 1.5)

We can also change the type of test performed (default: Wilcoxon test) or display the p-value as a symbol instead of a number.

ggplot(subset(mpg,class%in%c("suv","compact")),aes(x=class,y=cty))+
  geom_boxplot(notch = TRUE,width=0.6)+theme_bw()+
  stat_compare_means(label.x = 1.5,method = "t.test")

ggplot(subset(mpg,class%in%c("suv","compact")),aes(x=class,y=cty))+
  geom_boxplot(notch = TRUE,width=0.6)+theme_bw()+
  stat_compare_means(label.x = 1.5,method = "t.test",aes(label = ..p.signif..))



Comparing the means of multiple independent groups

If we do not have only two groups but more, we have to decide which comparisons we want to see. For example:

my_comparisons=list(c("f", "4"), c("4", "r"), c("r", "f"))
ggplot(mpg, aes(x=drv, y=cty))+geom_boxplot()+theme_bw()+stat_compare_means(comparisons = my_comparisons)

my_comparisons=list(c("f", "4"), c("4", "r"), c("r", "f"))
ggplot(mpg, aes(x=drv, y=cty))+geom_boxplot()+theme_bw()+stat_compare_means(comparisons = my_comparisons,aes(label = ..p.signif..))

This means that there is statistical difference between “4” and “f” and between “r” and “f”, but there is not between “4” and “r”.



Table plot

Sometimes you have to show a data table as image. ggpubr allows you to do that through the ggtexttable() function.

You can simply give your dataframe to the function, but you can also customize your plot by defining theme, styles, colors.

Visit https://rpkgs.datanovia.com/ggpubr/reference/ggtexttable.html for all options. You will find lots of examples.

Let’s create a dataframe and make a figure with it.

df=cbind.data.frame(flower=c("rose","daffodil", "tulip", "orchid", "lily"), N_remaining=c(5,20,14,2,0))
df
##     flower N_remaining
## 1     rose           5
## 2 daffodil          20
## 3    tulip          14
## 4   orchid           2
## 5     lily           0
ggtexttable(df, rows = NULL)

ggtexttable(df, rows = NULL, theme = ttheme("classic"))

ggtexttable(df, rows = NULL,
           theme = ttheme(
             colnames.style = colnames_style(color = "white", fill = "#8cc257"),
             tbody.style = tbody_style(color = "black",fill = c("#e8f3de", "#d3e8bb"))
           )
)



Multiple ggplots on the same page

ggarrange() function can arrange toghether multiple ggplot objects and, if you want, create a unique legend. You can also specify how many rows or columns do you prefer.

a=ggplot(mpg, aes(x=hwy, y=cty, color=class))+geom_point()+theme_bw()
b=ggplot(mpg, aes(x=as.character(year), y=cty))+geom_boxplot()+geom_jitter(aes(color=class))+theme_bw()+xlab("Years")
ggarrange(a,b, nrow = 1, common.legend = TRUE)



Publication-ready plots

Exploiting ggpubr package more in detail we can create a ready-to-publish figure that combines multiple ggplots on the same page or over multiple pages.

  • Create some graphs
a=ggplot(mpg, aes(x=hwy, y=cty, color=class))+geom_point()+theme_bw()
b=ggplot(mpg, aes(x=as.character(year), y=cty))+geom_boxplot()+geom_jitter(aes(color=class))+theme_bw()+xlab("Years")

df2=as.data.frame(table(mpg$year, mpg$class))
colnames(df2)=c("year", "class", "N")

c=ggtexttable(df2, rows = NULL, theme = ttheme("classic"))
d=ggplot(mpg, aes(x=hwy, y=cty))+geom_density2d_filled()+theme_bw()+ggtitle("Density 2D")
  • Arrange them in a unique panel and add labels
ggarrange(a,b,c,d, labels = c("A", "B", "C", "D"))

ggarrange(a,b,c,d, labels = c("A", "B", "C", "D"), 
          font.label=list(size = 20, color = "grey50", face = "bold"),
          hjust=-1, # Adjusts the horizontal position of each label. More negative values move the label further to the right on the plot canvas.
          vjust=1 # Adjusts the vertical position of each label. More positive values move the label further down on the plot canvas.
          )

# if we try using common.legend=T we see that the legend of density plot disappears.
  • We can also define widths and heights of columns in a grid (using relative numbers).
ggarrange(a,b,c,d, labels = c("A", "B", "C", "D"), 
          widths = c(1,2), # the right column is double the first
          heights=c(1, 1.5) # the second row is 1.5 times the first
          )

  • Add annotations. Note that the function annotate_figure() supports any ggplots.
figure=ggarrange(a,b,c,d, labels = c("A", "B", "C", "D"),
          widths = c(1,2), # the right column is double the first
          heights=c(1, 1.5) # the second row is 1.5 times the first
          ) 

figure=annotate_figure(figure,
                top = text_grob("Visualizing mpg", color = "magenta", face = "bold", size = 14),
                bottom = text_grob("Data source: \n mtcars data set", color = "navyblue",
                                   hjust = 1, x = 1, face = "italic", size = 10),
                left = text_grob("Figure arranged using ggarrange() function", color = "green", rot = 90),# rot indicate rotation angle
                right = "Figure completed",
                fig.lab = "Figure 1", fig.lab.face = "bold"
                )

figure

  • Align plot panels. We can specify whether graphs in the grid should be horizontally (h) or vertically (v) aligned; options are none (default), hv (align in both directions), h, and v.
ggarrange(a,b,c,d, labels = c("A", "B", "C", "D"), align = "v",legend=c("none", "right", "none", "right"))

  • Nested ggarrange. Maybe we want to arrage plots in a way that different rows contain different number of plots.
plot1=ggarrange(a,b,c, labels = c("A", "B", "C"), nrow=1, common.legend = T) # Put "a", "b" and "c" on a unique line; there is only one legend that is repeated, so we can use common.legend to unify it

ggarrange(plot1,d, labels=c("","D") , nrow=2)