(R) geom_violin() - a density graph

<geom_violin>

geom_violin is an efficient density graph. 

The violin graph is created by mirroring the kernel density curve symmetrically. 

It is usefully used for comparing several density graphs without overlapping. 

geom_violin()	geom_density()
> ggplot(data=birthwt,aes(x=smoke,y=bwt))+geom_violin()	> ggplot(data=birthwt,aes(x=bwt))+geom_density()+facet_grid(vars(smoke)) >ggplot(data=birthwt,aes(x=bwt,colour=smoke))+geom_density()

library(MASS) >install.packages(plyr) > table(birthwt$race)  1  2  3  96 26 67  > birthwt$race<-factor(birthwt$race) > birthwt$race<-revalue(birthwt$race,c("1"="white","2"="black","3"="others")) > table(birthwt$race)  white  black others      96     26     67

library(MASS) >install.packages(plyr) > table(birthwt$smoke)   0   1  115  74  > birthwt$smoke<-factor(birthwt$smoke) > birthwt$smoke<-revalue(birthwt$smoke,c("0"="No smoke","1"="smoke")) > table(birthwt$smoke) No smoke    smoke       115       74

bwt: birth weight in grams. 

> ggplot(data=birthwt,aes(x=smoke,y=bwt))+geom_violin()	> ggplot(data=birthwt,aes(x=smoke,y=bwt))+geom_violin(trim=FALSE)

> ggplot(data=birthwt,aes(x=smoke,y=lwt))+geom_violin()+geom_boxplot(fill="black")

You should take account that geom_violin() must be followed by geom_boxplot() if you want to express both graphs. 

Otherwise, the violin graph will cover the boxplot. 

> ggplot(data=birthwt,aes(x=smoke,y=lwt))+geom_violin()+geom_boxplot(fill="black",width=0.1)

> ggplot(data=birthwt,aes(x=smoke,y=lwt))+geom_violin()+geom_boxplot(fill="black",width=0.1)+stat_summary(fun=median,geom="point",fill="white",shape=21,size=3)

<from density to count - geom_violin(scale="count")>

Violin graph is a great substitute for other density functions. 

However, it can also express count instead of density if you set 'scale="count"' in geom_violin()

> ggplot(data=birthwt,aes(x=smoke,y=bwt))+geom_violin(scale="count")

<The way to adjust smoothness>

As you can see below, whether having adjust=1 or not does not affect the smoothness of the graph. 

It means that adjust for the density graph is originally set as 1. 

> ggplot(data=birthwt,aes(x=smoke,y=bwt))+geom_violin(adjust=1)	> ggplot(data=birthwt,aes(x=smoke,y=bwt))+geom_violin()

> ggplot(data=birthwt,aes(x=smoke,y=bwt)) +geom_violin(adjust=0.5)	> ggplot(data=birthwt,aes(x=smoke,y=bwt)) +geom_violin()	> ggplot(data=birthwt,aes(x=smoke,y=bwt)) +geom_violin(adjust=2)