英文:
Coloring by groups using ggscatter
问题
I'm trying to color my dots on this plot:
我想要给这个图上的点着色：
q <- ggscatter(ARAL_R, x = "Intl1", y = "Sul1", add = "reg.line", conf.int = TRUE, cor.coef = TRUE, cor.method = "spearman", xlab = "Intl1 (copies/g)", ylab = "Sul1 (copies/g)") + stat_cor(label.y = 1, label.x.npc = "center") + stat_regline_equation(label.y = 0.5, label.x.npc = "center")
但是每当我添加 color = "Core" 或 fill = "Core" 命令时（其中 A 是一个有 5 个类别的变量），就会出现以下情况：
但是每当我添加 color = "Core" 或 fill = "Core" 命令时（其中 A 是一个有 5 个类别的变量），就会出现以下情况：
q <- ggscatter(ARAL_R, x = "Intl1", y = "Sul1", color = "Core", add = "reg.line", conf.int = TRUE, cor.coef = TRUE, cor.method = "spearman", xlab = "Intl1 (copies/g)", ylab = "Sul1 (copies/g)") + stat_cor(label.y = 1, label.x.npc = "center") + stat_regline_equation(label.y = 0.5, label.x.npc = "center")
ggpar(q, xscale = "log10", yscale = "log10")
我想要它像这样：
我想要它像这样：
But with the spearman's correlation line.
但是带有 Spearman 相关性线。
What am I doing wrong?
我做错了什么？
英文:
I'm trying to color my dots on this plot:
q <- ggscatter(ARAL_R, x = "Intl1", y = "Sul1", add = "reg.line", conf.int = TRUE, cor.coef = TRUE, cor.method = "spearman", xlab = "Intl1 (copies/g)", ylab = "Sul1 (copies/g)") + stat_cor(label.y = 1, label.x.npc = "center") + stat_regline_equation(label.y = 0.5, label.x.npc = "center")
But whenever I add the command color = "Core" or fill = "Core" (where A is a categorical variable of 5) this happens:
q <- ggscatter(ARAL_R, x = "Intl1", y = "Sul1", color = "Core", add = "reg.line", conf.int = TRUE, cor.coef = TRUE, cor.method = "spearman", xlab = "Intl1 (copies/g)", ylab = "Sul1 (copies/g)") + stat_cor(label.y = 1, label.x.npc = "center") + stat_regline_equation(label.y = 0.5, label.x.npc = "center")ggpar(q, xscale = "log10", yscale = "log10")
I would like to make it like this:
But with the spearman's correlation line.
What am I doing wrong?
答案1
得分: 0
以下是您要翻译的内容：
问题在于当将变量映射到 color 上时，您的数据会分成多个组，并且您会获得每个 Core 组或类别的回归线、相关系数和方程式。要解决这个问题，您必须显式地在 group aes 上进行映射，即使用 aes(group = 1) 来将所有观测值视为一个组，通常我们将其命名为 1，这是为了简化或按照惯例。不幸的是，ggscatter 不提供这个选项。这是使用 "开箱即用" 选项而不是使用原始的 ggplot2 时的一个缺点之一。但我担心这是前进的方式，即使用 geom_smooth 手动添加您的回归线，而不依赖于 ggscatter：
使用基于 iris 的最小可复制示例：
library(ggpubr)#> Loading required package: ggplot2ARAL_R <- iris[3:5]names(ARAL_R) <- c("Intl1", "Sul1", "Core")q <- ggscatter(ARAL_R,x = "Intl1", y = "Sul1", color = "Core",xlab = "Intl1 (copies/g)", ylab = "Sul1 (copies/g)") +geom_smooth(aes(group = 1), method = "lm", color = "black") +stat_cor(aes(group = 1), label.y = 1, label.x.npc = "center") +stat_regline_equation(aes(group = 1), label.y = 0.5, label.x.npc = "center")ggpar(q, xscale = "log10", yscale = "log10")#> `geom_smooth()` using formula = 'y ~ x'
作为参考，这是原始 ggplot2 的方式，其中我将 group aes 替换为将 color 仅作为 geom_point 的本地 aes，并且仅使用 ggpubr 添加回归线方程和相关系数：
library(ggplot2)library(ggpubr)ARAL_R <- iris[3:5]names(ARAL_R) <- c("Intl1", "Sul1", "Core")ggplot(ARAL_R,aes(Intl1, Sul1),) +geom_smooth(method = "lm", color = "black") +geom_point(aes(color = Core)) +stat_cor(label.y = 1, label.x.npc = "center") +stat_regline_equation(label.y = 0.5, label.x.npc = "center") +labs(x = "Intl1 (copies/g)", y = "Sul1 (copies/g)") +scale_x_log10() +scale_y_log10()#> `geom_smooth()` using formula = 'y ~ x'
英文:
The issue is that when mapping a variable on color your data gets split into multiple groups and you get a regression line, correlation coefficient and equation for each group or category of Core. To fix that you have to explicitly map on the group aes, i.e. use aes(group = 1) to treat all obs. as one group which for simplicity or by convention we name 1. Unfortunately ggscatter does not offer this option. That's one of the downsides when using "out-of-the-box" options instead of using vanilla ggplot2. But I'm afraid this is the way to go, i.e. add your regression line manually using a geom_smooth instead of relying on ggscatter:
Using a minimal reproducible example based on iris:
library(ggpubr)#> Loading required package: ggplot2ARAL_R <- iris[3:5]names(ARAL_R) <- c("Intl1", "Sul1", "Core")q <- ggscatter(ARAL_R,x = "Intl1", y = "Sul1", color = "Core",xlab = "Intl1 (copies/g)", ylab = "Sul1 (copies/g)") +geom_smooth(aes(group = 1), method = "lm", color = "black") +stat_cor(aes(group = 1), label.y = 1, label.x.npc = "center") +stat_regline_equation(aes(group = 1), label.y = 0.5, label.x.npc = "center")ggpar(q, xscale = "log10", yscale = "log10")#> `geom_smooth()` using formula = 'y ~ x'
<!-- -->
And as a reference here is the vanilla ggplot2 way, where instead of using the group aes I made color a local aes of the geom_point only and where I use ggpubr only to add the regline equation and correlation coefficient:
library(ggplot2)library(ggpubr)ARAL_R <- iris[3:5]names(ARAL_R) <- c("Intl1", "Sul1", "Core")ggplot(ARAL_R,aes(Intl1, Sul1),) +geom_smooth(method = "lm", color = "black") +geom_point(aes(color = Core)) +stat_cor(label.y = 1, label.x.npc = "center") +stat_regline_equation(label.y = 0.5, label.x.npc = "center") +labs(x = "Intl1 (copies/g)", y = "Sul1 (copies/g)") +scale_x_log10() +scale_y_log10()#> `geom_smooth()` using formula = 'y ~ x'
<!-- -->
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论