我有使用距离矩阵上的cmdscale命令提供或生成的p值，x值和y值向量。只需使用plot(x,y)绘制坐标即可，但我希望将点的大小与其p值成比例(较小的p值表示较大的点)。我不太想办法做到这一点，并且正在寻找建议。我曾想过要通过一些因素(plot(..., cex=2*normalized))对p值进行标准化和缩放，但这是行不通的。下面，我转储了一些我正在使用的示例值，以实现可重复性。
> dput(pValues)
c(4.48e-14, 1.66e-12, 2.53e-08, 8.57e-08, 3.4e-07, 5.68e-07,
9.92e-07, 1.08e-06, 2.82e-06, 1.81e-05, 0.000133, 0.00053, 0.000616,
0.000846, 0.000947, 0.001110537, 0.001110537, 0.001505779, 0.001573054,
0.001573054, 0.002112306, 0.002308863, 0.003121497, 0.003121497,
0.003121497, 0.003121497, 0.003121497, 0.003121497, 0.003121497,
0.003121497, 0.003177736, 0.004723347, 0.005004768, 0.005301549,
1.86e-17, 9.18e-17, 2.16e-16, 8.23e-16, 9.2e-16, 1.28e-15, 1.38e-15,
2.59e-15, 6.43e-15, 6.43e-15, 8.42e-15, 1.21e-14, 1.02e-13, 7.58e-13,
1.53e-12, 1.96e-11)
> dput(x)
c(-0.546606289027691, -0.513646680083475, 0.157100976250898,
0.109447441578375, 0.109447441578375, 0.104451507558839, 0.104451507558839,
0.109447441578375, 0.175507893375115, -0.14664445744836, 0.0543475836486623,
0.0557408040609083, 0.0893466913878634, 0.0893466913878634, 0.142438485025367,
0.0470980043880961, -0.0221917747418056, 0.109447441578375, 0.0362416205348296,
0.0470980043880961, 0.0362416205348296, 0.0347865097394601, 0.0391497309324339,
0.0413674642703439, 0.0667384023198892, 0.0461182424640277, 0.0413674642703439,
0.0667384023198892, 0.0461182424640277, 0.0475891023261346, 0.0893466913878634,
0.0764742527259463, 0.0422421029990655, -0.0221917747418056,
-0.510082195428624, -0.510082195428624, -0.510082195428624, -0.510082195428624,
0.53984552027647, 0.457352428403424, -0.510082195428624, -0.510082195428624,
0.476216399097293, 0.476216399097293, -0.510082195428624, 0.297997535161347,
-0.510082195428624, 0.397117197655551, 0.440730282360781, 0.0312250127868402)
> dput(y)
c(0.107461316099316, 0.156755909792581, -0.166842986685387,
-0.141978234324384, -0.141978234324384, -0.0687959347159215,
-0.0687959347159215, -0.141978234324384, -0.142554658469002,
-0.0395153544691704, -0.0576565915449701, -0.0936541502757846,
-0.0438034590304964, -0.0438034590304964, -0.190330058396921,
-0.0329359077881266, -0.0116066646384657, -0.141978234324384,
-0.0714188307783769, -0.0329359077881266, -0.0714188307783769,
-0.054867626805721, -0.0112558858117774, -0.0166800568953671,
-0.0274480805166001, -0.0331407851151761, -0.0166800568953671,
-0.0274480805166001, -0.0331407851151761, -0.00455654056913195,
-0.0438034590304963, -0.0148236474766705, -0.130181815402346,
-0.0116066646384657, 0.0838569446695995, 0.0838569446695995,
0.0838569446695995, 0.0838569446695995, 0.0372937912551249, 0.555328846358372,
0.0838569446695995, 0.0838569446695995, 0.521415820920117, 0.521415820920117,
0.0838569446695994, -0.506985517718071, 0.0838569446695995, -0.324019743520653,
0.421305271998988, -0.0312119222707089)
最佳答案
您可以将它们绑定(bind)到data.frame并对其进行ggplot:
df=data.frame(x,y,pValues)
library(ggplot2)
ggplot(data=df) + aes(x=x, y=y, size=-log(pValues)) + geom_point(alpha=0.5, col='blue')
我建议直接绘制p值的对数，取相反的值，这样您将获得正确的直观方式(越大，则越有意义)
这是快速的方法。如果要自定义图并改善图例，我们可以在scale_size的
trans参数中直接指定对数变换。您还可以弄乱范围(圆的大小范围)，将在图例中使用的中断(请小心，使用原始单位)，甚至图例标题。ggplot(data=df) + aes(x=x, y=y, size=pValues) + geom_point(alpha=0.5, col='blue') +
scale_size("p-values", trans="log10", range=c(15, 1), breaks=c(1e-17, 1e-15, 1e-10, 1e-5, 1e-3))
请注意，由于转换函数中没有减号，因此我必须反转范围限制的顺序。
关于r - 如何使图上的点的大小与p值成比例？，我们在Stack Overflow上找到一个类似的问题：https://stackoverflow.com/questions/38310192/