r - 绘制修改后的点和线图 - 可变性为 r 中的“尖峰”图

Question

在解释细节之前，这是我的数据：

set.seed (1234) 
datas <- data.frame (Indv = 1:20, Xvar = rnorm (20, 50, 10),
Yvar = rnorm (20, 30,5), Yvar1 = rnorm (20, 10, 2),
Yvar2 = rnorm (20, 5, 1), Yvar3 = rnorm (20, 100, 20),
Yvar4 = rnorm (20, 15, 3))

我想准备一个图（Metroglymph），它本质上是点图，但是（Xvar 和 Yvar 的）点（Xvar 和 Yvar）具有从缩放到其余变量（Yvar1、Yvar2、Yvar3、Yvar4）的点的尖峰（线）。每个钉子都是有序的，最好用颜色编码。

require(ggplot2)
ggplot(datas, aes(x=Xvar, y=Yvar)) +
    geom_point(shape=1, size = 10) + theme_bw()

Answer 1

这是一种可能对您有帮助的方法。它使用stat_spoke()来自 ggplot2。您的每个 y 变量都在 4 次单独调用中映射到辐条半径stat_spoke。

plot_1 = ggplot(datas, aes(x=Xvar, y=Yvar)) +
         stat_spoke(aes(angle=(1/8)*pi, radius=Yvar1), colour="#E41A1C",size=1) +
         stat_spoke(aes(angle=(3/8)*pi, radius=Yvar2), colour="#377EB8",size=1) +
         stat_spoke(aes(angle=(5/8)*pi, radius=Yvar3), colour="#4DAF4A",size=1) +
         stat_spoke(aes(angle=(7/8)*pi, radius=Yvar4), colour="#984EA3",size=1) +
         geom_point(shape=1, size = 10)

ggsave("plot_1.png", plot_1)

根据您的数据和特定需求，转换变量以使其更适合绘图可能是有意义的。

normalize = function(x) {
    new_x = (x - mean(x)) / sd(x)
    new_x = new_x + abs(min(new_x))
    return(new_x)
}

plot_2 = ggplot(datas, aes(x=Xvar, y=Yvar)) +
         stat_spoke(aes(angle=(1/8)*pi, radius=normalize(Yvar1)), colour="#E41A1C", size=1) +
         stat_spoke(aes(angle=(3/8)*pi, radius=normalize(Yvar2)), colour="#377EB8", size=1) +
         stat_spoke(aes(angle=(5/8)*pi, radius=normalize(Yvar3)), colour="#4DAF4A", size=1) +
         stat_spoke(aes(angle=(7/8)*pi, radius=normalize(Yvar4)), colour="#984EA3", size=1) +
         geom_point(shape=1, size = 10)

ggsave("plot_2.png", plot_2)

重要警告：对于相同的辐条半径值，如果线更垂直，则绘制线的幅度会更大，如果线更水平，则绘制线的幅度会更小。这是因为 x 的范围大约是数据集 y 范围的两倍。随着 x 与 y 轴比的变化，绘制的角度也变得扭曲。添加coord_equal(ratio=1)解决了这个问题，但可能会引入其他问题。

编辑：无循环绘图

这很有趣，也很有教育意义。键入重复的代码可能会更省时！如果有人可以提供改进此代码的建议，请发表评论。

library(reshape2)

dat2 = melt(datas, id.vars=c("Indv", "Xvar", "Yvar"), 
            variable.name="spoke_var", value.name="spoke_value")

# Apply normalization in a loop. Can plyr do this more gracefully?.
for (var_name in levels(dat2$spoke_var)) {
    select_rows = dat2$spoke_var == var_name
    norm_dat = normalize(dat2[select_rows, "spoke_value"])
    dat2[select_rows, "spoke_value"] = norm_dat
}

# Pick an angle for each Yvar, then add angle column to dat2.
tmp = data.frame(spoke_var=unique(dat2$spoke_var))
tmp$spoke_angle = seq(from=pi/8, by=pi/4, length.out=nrow(tmp))
dat2 = merge(dat2, tmp)

plot_4 = ggplot(dat2, aes(x=Xvar, y=Yvar)) +
         stat_spoke(data=dat2, size=1,
                    aes(colour=spoke_var, angle=spoke_angle, radius=spoke_value)) +
         geom_point(data=datas, aes(x=Xvar, y=Yvar), shape=1, size=7) +
         coord_equal(ratio=1) +
         scale_colour_brewer(palette="Set1")

Answer 2

这是更多手动方法：

set.seed (1234)
datas <- data.frame (Indv = 1:20, Xvar = rnorm (20, 50, 10),
Yvar = rnorm (20, 30,5), Yvar1 = rnorm (20, 10, 2),
Yvar2 = rnorm (20, 5, 1), Yvar3 = rnorm (20, 100, 20),
Yvar4 = rnorm (20, 15, 3))
datas$SYvar1 <- 2 + scale (datas$Yvar1)
datas$SYvar2 <- 2 +  scale (datas$Yvar2)
datas$SYvar3 <- 2 + scale (datas$Yvar3)
datas$SYvar4 <- 2 +  scale (datas$Yvar4)

    require(ggplot2)
    p <- ggplot(datas, aes(x=Xvar, y=Yvar)) +  
   geom_point(size = 10, pch = 19, col = "yellow2")
    p + geom_segment(aes(x = Xvar, y = Yvar, xend = Xvar + SYvar1, 
   yend = Yvar), col = "red4", size = 1) +
    geom_segment(aes(x = Xvar, y = Yvar, xend = Xvar, 
    yend = Yvar + SYvar2), col = "green4", size = 1) +
    geom_segment(aes(x = Xvar, y = Yvar, xend = Xvar-2.5,
    yend = Yvar + SYvar3), col = "darkblue", size = 1) +
    geom_segment(aes(x = Xvar, y = Yvar, xend = 
     Xvar - SYvar4, yend = Yvar ), col = "red", size = 1) +
    theme_bw()