使用R中的ggtern进行三元图的凸壳

时间:2017-08-28 16:27:43

标签: r ggplot2 ggtern

多边形凸壳区域

考虑以下data.frame

DGChi <- structure(list(Sucrose = c(42, 40, 15, 19, 33, 49, 35, 31, 22, 
25, 37, 28, 31, 41, 27, 28, 33, 43, 21, 37, 14, 41, 30, 34, 38, 
40, 40, 33, 33), Fructose = c(27, 29, 41, 35, 29, 23, 27, 33, 
38, 38, 28, 31, 29, 26, 32, 34, 31, 28, 40, 30, 39, 27, 32, 31, 
29, 28, 28, 32, 29), Glucose = c(31, 31, 44, 46, 38, 28, 38, 
36, 40, 37, 35, 41, 40, 33, 41, 38, 36, 30, 39, 33, 47, 32, 38, 
35, 33, 32, 32, 35, 38), Sindrome = c("Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily")), .Names = c("Sucrose", 
"Fructose", "Glucose", "Sindrome"), row.names = c(NA, -29L), class = c("tbl_df", 
"tbl", "data.frame"))

首先尝试

我正在尝试制作三元图并在点周围添加凸包,我的第一次尝试是使用ggalt包的 geom_encircle

library(ggtern)
library(ggalt)

ggtern(data = DGChi, aes(x = Fructose, y = Sucrose, z = Glucose, fill = Sindrome)) +
theme_bw() +
geom_encircle(alpha=0.2,size=1, spread = 0.5) +
geom_point() +
theme(legend.position="bottom") 

使用此结果

enter image description here

围绕点,但不是凸包

第二次尝试使用几何包

尝试关注my own answer for rgl,我尝试了这个:

library(geometry)
DGChiMin <- as.data.frame(convhulln(matrix(c(DGChi$Fructose, DGChi$Sucrose, DGChi$Glucose), ncol = 3)))
colnames(DGChiMin) <- c("Fructose", "Sucrose", "Glucose")

然后是剧情:

ggtern(data = DGChi, aes(x = Fructose, y = Sucrose, z = Glucose)) +
  theme_bw() +
  geom_polygon(data = DGChiMin,aes(x = Fructose, y = Sucrose, z = Glucose)) +
  geom_point() +
  theme(legend.position="bottom")

但得到了这个超级奇怪的多边形:

enter image description here

有人可以帮助我获得凸壳图吗?

2 个答案:

答案 0 :(得分:4)

认为你会发现geom_encircle在内部使用chull。将expand参数设置为0。

library(ggalt)
library(ggtern)
ggtern(data = DGChi, aes(x = Fructose, y = Sucrose, z = Glucose, fill = Sindrome)) +
  theme_bw() +
  geom_encircle(alpha=0.2,size=1, expand=0) + ##<<<<<< expand = 0
  geom_point() +
  theme(legend.position="bottom") 

Output

答案 1 :(得分:2)

以下是基于chull函数的简单解决方案。

DGChi <- structure(list(Sucrose = c(42, 40, 15, 19, 33, 49, 35, 31, 22, 
25, 37, 28, 31, 41, 27, 28, 33, 43, 21, 37, 14, 41, 30, 34, 38, 
40, 40, 33, 33), Fructose = c(27, 29, 41, 35, 29, 23, 27, 33, 
38, 38, 28, 31, 29, 26, 32, 34, 31, 28, 40, 30, 39, 27, 32, 31, 
29, 28, 28, 32, 29), Glucose = c(31, 31, 44, 46, 38, 28, 38, 
36, 40, 37, 35, 41, 40, 33, 41, 38, 36, 30, 39, 33, 47, 32, 38, 
35, 33, 32, 32, 35, 38), Sindrome = c("Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily", "Chiropterophily", 
"Chiropterophily", "Chiropterophily", "Chiropterophily")), .Names = c("Sucrose", 
"Fructose", "Glucose", "Sindrome"), row.names = c(NA, -29L), class = c("tbl_df", 
"tbl", "data.frame"))

# Convex hull
ch <- chull(DGChi[,1:2])
DGChiMin <- DGChi[ch, 1:3]

library(ggtern)
ggtern(data = DGChi, aes(x = Fructose, y = Sucrose, z = Glucose)) + geom_point() +
  theme_bw() +
  geom_polygon(data = DGChiMin, aes(x = Fructose, y = Sucrose, z = Glucose), fill="#FF000044") +
  theme(legend.position="bottom")

enter image description here