在 size legend 中显示所有形状 geom_point ggplot2

Question

我需要在我的图表中使用三种形状，但是只有实心圆被用于尺寸图例（显示在图表上）。我正在尝试找到一种方法来在该图例中包含所有的形状，是否有办法做到呢？

数据框和绘图代码：

data <- structure(list(date = structure(c(1431388800, 1444780800, 1456876800, 
    1469145600, 1469664000, 1425081600, 1445299200, 1488758400, 1524960000, 
    1454544000, 1540512000, 1429228800, 1481587200, 1526688000, 1445904000, 
    1498348800, 1464825600, 1451174400, 1450310400, 1539216000, 1528934400, 
    1525996800, 1465084800, 1531180800), class = c("POSIXct", "POSIXt"
    ), tzone = "UTC"), cnt1 = c(76, 140, 99, 10, 123, 124, 160, 58, 
    47, 43, 86, 112, 26, 123, 111, 49, 65, 93, 174, 49, 5, 16, 84, 
    4), cnt2 = c(111, 200, 550, 73, 123, 184, 166, 131, 189, 136, 
    250, 195, 26, 154, 125, 100, 168, 151, 255, 49, 200, 144, 177, 
    4), per = c(68.4684684684685, 70, 18, 13.6986301369863, 100, 
    67.3913043478261, 96.3855421686747, 44.2748091603053, 24.8677248677249, 
    31.6176470588235, 34.4, 57.4358974358974, 100, 79.8701298701299, 
    88.8, 49, 38.6904761904762, 61.5894039735099, 68.2352941176471, 
    100, 2.5, 11.1111111111111, 47.4576271186441, 100), status = c("A", 
    "B", "C", "C", "B", "B", "B", "A", "A", "A", "C", "C", "C", "A", 
    "B", "C", "B", "C", "A", "A", "A", "B", "B", "B"), category = c("CAT3", 
    "CAT1", "CAT3", "CAT2", "CAT1", "CAT1", "CAT1", "CAT3", "CAT2", 
    "CAT3", "CAT2", "CAT1", "CAT1", "CAT1", "CAT2", "CAT2", "CAT1", 
    "CAT1", "CAT1", "CAT1", "CAT3", "CAT3", "CAT3", "CAT1")), .Names = c("date", 
    "cnt1", "cnt2", "per", "status", "category"), row.names = c(NA, 
    -24L), class = c("tbl_df", "tbl", "data.frame"))



 ggplot(data, aes(x=date, y=cnt2)) + 
      scale_y_log10(breaks=c(1,10,100,1000,5000)) +
      scale_size_continuous(trans="sqrt",range = c(1, 5),breaks=c(5,10,25,50,75,100)) +
      geom_point(aes(col=category, size=per, shape=status), alpha=0.7)  

- Ibo

---

看起来你可以使用override.aes（在scale_size_*中添加guide = guide_legend(override.aes = list(shape = 17))）更改大小图例中的形状，但这并没有回答你的问题。你想让这样的图例是什么样子？你想象中有多列吗？ - aosmith

@aosmith 基本上理想情况是在图例的每一行中，一个圆形、一个三角形和一个正方形相邻。这样更容易知道所有形状的大小。如果没有视觉参考，看圆形并猜测同样大小的三角形会是什么样子是很困难的。 - Ibo

Answer 1

我找到了一个解决办法，因为似乎没有简单的方法来做到这一点。理想的解决方案是将形状放在一起，并在图例中只提及数字而不是重复它。此外，这个解决方法可以更加参数化，因为通过改变status的数量并显示一个新的形状，所有输入覆盖函数的向量都可以正确构造：

    l=c(5,10,25,50,75,100)
    mybreaks=c(rep(l,3))
    myshapes=c(rep(15,6),rep(16,6),rep(17,6))

    ggplot(data, aes(x=date, y=cnt2)) + 
    scale_y_log10(breaks=c(1,10,100,1000,5000)) +
    scale_size_continuous(trans="sqrt",range = c(1, 5),breaks=mybreaks) +
    geom_point(aes(col=category, size=per, shape=status), alpha=0.7)  +
    guides(size=guide_legend(ncol=3,override.aes = list( shape =myshapes)))

- Ibo