在 ggplot2 中添加完全相同数量的水平扩展/“抖动”/闪避答案

【问题标题】：在 ggplot2 中添加完全相同数量的水平扩展/“抖动”/闪避
【发布时间】：2022-01-22 19:19:35
【问题描述】：

我想绘制三个组的测量值，每个组有五个观察值。我想绘制所有点，但是每组中的数值非常接近。更改 alpha 会有所帮助，但仍然很难观察到单独的点。

因此，我想添加一些水平扩展（在 X 轴上）。点的部分重叠是可以的。 geom_jitter() 可以实现这一点，但会增加随机分布，而我想在 X 轴上均匀分布五个点，因为水平轴上的数据没有随机性。

Geom_dotplot() 可以创建均匀分布，但据我了解，仅适用于离散/分箱值。

这是我的 ggplot 输出，以及带有 tribble 格式数据和 ggplot 代码的可重现代码：





library(tidyverse)

df <- tribble(
  ~projection,               ~E.DAP, ~Shielding,
  "AP", 2.88755980861244e-06,     "None",
  "AP", 2.87626262626263e-06,     "None",
  "AP",             2.87e-06,     "None",
  "AP", 2.87702265372168e-06,     "None",
  "AP", 2.87775551102204e-06,     "None",
  "CRAN", 2.93893129770992e-06,     "None",
  "CRAN", 2.92857142857143e-06,     "None",
  "CRAN", 2.93517017828201e-06,     "None",
  "CRAN",           2.9375e-06,     "None",
  "CRAN", 2.93831168831169e-06,     "None",
  "LAO", 6.88053097345133e-06,     "None",
  "LAO", 6.90217391304348e-06,     "None",
  "LAO", 6.84489795918367e-06,     "None",
  "LAO", 6.84792626728111e-06,     "None",
  "LAO", 6.86425339366516e-06,     "None",
  "LAO-CRAN", 7.11216730038023e-06,     "None",
  "LAO-CRAN",  7.1039501039501e-06,     "None",
  "LAO-CRAN",  7.1042471042471e-06,     "None",
  "LAO-CRAN", 7.09368635437882e-06,     "None",
  "LAO-CRAN", 7.10183299389002e-06,     "None",
  "RAO",             2.48e-06,     "None",
  "RAO", 2.47540983606557e-06,     "None",
  "RAO", 2.46979865771812e-06,     "None",
  "RAO", 2.46206896551724e-06,     "None",
  "RAO",         2.453125e-06,     "None",
  "RAO-CRAN", 1.87972508591065e-06,     "None",
  "RAO-CRAN", 1.87068965517241e-06,     "None",
  "RAO-CRAN", 1.88461538461538e-06,     "None",
  "RAO-CRAN", 1.86816720257235e-06,     "None",
  "RAO-CRAN",  1.8735632183908e-06,     "None",
  "LAO-CAUD", 6.03324808184143e-06,     "None",
  "LAO-CAUD", 6.06502242152466e-06,     "None",
  "LAO-CAUD", 6.04785894206549e-06,     "None",
  "LAO-CAUD",  6.0402144772118e-06,     "None",
  "LAO-CAUD", 6.02949061662198e-06,     "None",
  "RAO-CAUD", 6.30573248407643e-07,     "None",
  "RAO-CAUD", 6.29834254143646e-07,     "None",
  "RAO-CAUD", 6.35467980295566e-07,     "None",
  "RAO-CAUD", 6.38297872340426e-07,     "None",
  "RAO-CAUD", 6.37755102040816e-07,     "None",
  "CAUD", 3.81597222222222e-06,     "None",
  "CAUD", 3.83083511777302e-06,     "None",
  "CAUD",  3.8330550918197e-06,     "None",
  "CAUD", 3.82936507936508e-06,     "None",
  "CAUD", 3.81614349775785e-06,     "None",
  "AP", 4.53580901856764e-07, "Standard",
  "AP", 4.46745562130178e-07, "Standard",
  "AP",  4.4973544973545e-07, "Standard",
  "AP", 4.44976076555024e-07, "Standard",
  "AP", 4.52380952380952e-07, "Standard",
  "CRAN", 1.20574162679426e-06, "Standard",
  "CRAN", 1.20113314447592e-06, "Standard",
  "CRAN", 1.19130434782609e-06, "Standard",
  "CRAN", 1.19349593495935e-06, "Standard",
  "CRAN", 1.19757575757576e-06, "Standard",
  "LAO", 1.50961538461538e-06, "Standard",
  "LAO", 1.50761421319797e-06, "Standard",
  "LAO", 1.51209677419355e-06, "Standard",
  "LAO", 1.52216748768473e-06, "Standard",
  "LAO", 1.51476793248945e-06, "Standard",
  "LAO-CRAN", 2.96213425129088e-06, "Standard",
  "LAO-CRAN", 2.95991561181435e-06, "Standard",
  "LAO-CRAN",             2.95e-06, "Standard",
  "LAO-CRAN", 2.95744680851064e-06, "Standard",
  "LAO-CRAN", 2.95266272189349e-06, "Standard",
  "RAO", 1.51515151515152e-07, "Standard",
  "RAO", 1.52173913043478e-07, "Standard",
  "RAO", 1.52866242038217e-07, "Standard",
  "RAO", 1.57575757575758e-07, "Standard",
  "RAO", 1.59420289855072e-07, "Standard",
  "RAO-CRAN", 2.41379310344828e-07, "Standard",
  "RAO-CRAN", 2.33050847457627e-07, "Standard",
  "RAO-CRAN", 2.34741784037559e-07, "Standard",
  "RAO-CRAN", 2.33812949640288e-07, "Standard",
  "RAO-CRAN", 2.34817813765182e-07, "Standard",
  "LAO-CAUD", 1.89125295508274e-07, "Standard",
  "LAO-CAUD", 1.87110187110187e-07, "Standard",
  "LAO-CAUD", 1.89309576837416e-07, "Standard",
  "LAO-CAUD", 1.90821256038647e-07, "Standard",
  "LAO-CAUD", 1.89189189189189e-07, "Standard",
  "AP",  3.7542662116041e-08,      "XRB",
  "AP", 3.89972144846797e-08,      "XRB",
  "AP", 4.15335463258786e-08,      "XRB",
  "AP", 3.83275261324042e-08,      "XRB",
  "AP", 4.09556313993174e-08,      "XRB",
  "CRAN",                4e-08,      "XRB",
  "CRAN", 4.08163265306122e-08,      "XRB",
  "CRAN", 4.06905055487053e-08,      "XRB",
  "CRAN", 3.94574599260173e-08,      "XRB",
  "CRAN", 3.90835579514825e-08,      "XRB",
  "LAO", 6.47249190938511e-08,      "XRB",
  "LAO", 6.69144981412639e-08,      "XRB",
  "LAO", 6.42570281124498e-08,      "XRB",
  "LAO", 6.19834710743802e-08,      "XRB",
  "LAO",             6.25e-08,      "XRB",
  "LAO-CRAN", 5.31914893617021e-08,      "XRB",
  "LAO-CRAN", 5.11247443762781e-08,      "XRB",
  "LAO-CRAN", 5.30821917808219e-08,      "XRB",
  "LAO-CRAN", 5.17857142857143e-08,      "XRB",
  "LAO-CRAN", 5.31732418524871e-08,      "XRB",
  "RAO", 3.30578512396694e-08,      "XRB",
  "RAO", 3.38983050847458e-08,      "XRB",
  "RAO", 3.33333333333333e-08,      "XRB",
  "RAO", 3.07017543859649e-08,      "XRB",
  "RAO",  3.2171581769437e-08,      "XRB",
  "RAO-CRAN", 3.44827586206897e-08,      "XRB",
  "RAO-CRAN",  3.5264483627204e-08,      "XRB",
  "RAO-CRAN", 3.25581395348837e-08,      "XRB",
  "RAO-CRAN", 3.43007915567282e-08,      "XRB",
  "RAO-CRAN", 3.37837837837838e-08,      "XRB",
  "LAO-CAUD", 5.70776255707763e-08,      "XRB",
  "LAO-CAUD", 5.50847457627119e-08,      "XRB",
  "LAO-CAUD", 5.51876379690949e-08,      "XRB",
  "LAO-CAUD", 5.42797494780793e-08,      "XRB",
  "LAO-CAUD", 5.53191489361702e-08,      "XRB",
  "RAO-CAUD",  2.1505376344086e-08,      "XRB",
  "RAO-CAUD", 1.52990264255911e-08,      "XRB",
  "RAO-CAUD", 1.58033362598771e-08,      "XRB",
  "RAO-CAUD", 1.47679324894515e-08,      "XRB",
  "RAO-CAUD", 1.60213618157543e-08,      "XRB",
  "CAUD", 3.84024577572965e-08,      "XRB",
  "CAUD", 3.98671096345515e-08,      "XRB",
  "CAUD",                4e-08,      "XRB",
  "CAUD", 3.92156862745098e-08,      "XRB",
  "CAUD", 3.94574599260173e-08,      "XRB",
  "RAO-CAUD", 4.52961672473868e-08, "Standard",
  "RAO-CAUD", 4.12844036697248e-08, "Standard",
  "RAO-CAUD", 4.83870967741936e-08, "Standard",
  "RAO-CAUD", 4.29184549356223e-08, "Standard",
  "RAO-CAUD", 4.38356164383562e-08, "Standard",
  "CAUD", 8.58974358974359e-08, "Standard",
  "CAUD", 8.70646766169154e-08, "Standard",
  "CAUD", 8.66372980910426e-08, "Standard",
  "CAUD", 8.67579908675799e-08, "Standard",
  "CAUD", 8.70113493064313e-08, "Standard",
  "LAO90", 3.38266384778013e-08,      "XRB",
  "LAO90", 3.34346504559271e-08,      "XRB",
  "LAO90",  3.2258064516129e-08,      "XRB",
  "LAO90", 3.21543408360129e-08,      "XRB",
  "LAO90", 3.20987654320988e-08,      "XRB",
  "LAO90", 8.60215053763441e-08, "Standard",
  "LAO90", 8.88888888888889e-08, "Standard",
  "LAO90", 8.44686648501362e-08, "Standard",
  "LAO90", 8.90804597701149e-08, "Standard",
  "LAO90", 8.67052023121387e-08, "Standard",
  "LAO90", 1.96982378854626e-05,     "None",
  "LAO90", 1.97443820224719e-05,     "None",
  "LAO90", 1.96883720930233e-05,     "None",
  "LAO90", 1.96856540084388e-05,     "None",
  "LAO90", 1.96415770609319e-05,     "None"
)

ggplot(df , aes(x=projection, y=E.DAP, color=Shielding)) + 
  geom_jitter(size = 2.5, width=0.2, height=0, alpha =0.8) + 
  theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1, size=8)) +
  labs(title= "mSv/DAP according to projection",
       y="mSv/mGy*cm2 ", x="")+
  theme(legend.position = c(0.85, 0.8), legend.key = element_rect(colour = "transparent", fill = "transparent")) +  
  scale_color_manual(values=c(c("#69ac5c", "#5b86c3","#c7533b")))  +
  theme(legend.title = element_blank())

【问题讨论】：

Shape=4 确实提高了可读性，但不会使水平展开对称。

标签： r ggplot2

【解决方案1】：

可以使用包ggbeeswarm... 外观可以通过不同的形状或geom_beeswarm 的参数来增强：alpha、dodge.width 或groupOnX。

设置dodge.width 强制将点按Shielding 变量分组。更新后的答案，名义上包括 0.1 的值，将 dodge.width 设置为更接近于 0 使得看起来好像没有闪避，并且将来自相同 Shielding 值的点保持在一起。较大的 dodge.width 值将有助于区分 x 变量的点，例如“CAUD”。

为确保Shielding 的重叠绘制是一致的，请创建一个绘图顺序变量。通过屏蔽排列数据任意设置，以便最后一个变量绘制最后，这可以改变以适应。

library(ggplot2)
library(dplyr)
library(ggbeeswarm)


df1 <- 
  df %>%  
  arrange(Shielding) %>% 
  mutate(plot_order = row_number())

ggplot(df1, aes(x = projection, y = E.DAP, color = Shielding, order = plot_order)) + 
  geom_beeswarm(size = 2.5, alpha = 0.9, dodge.width = 0.1) +
  scale_color_manual(values=c(c("#69ac5c", "#5b86c3","#c7533b")))  +
  labs(title= "mSv/DAP according to projection",
       y="mSv/mGy*cm2 ", x="")+
  theme(legend.title = element_blank(),
        legend.position = c(0.85, 0.8),
        legend.key = element_rect(colour = "transparent", fill = "transparent"),
        axis.text.x = element_text(angle = 45, vjust = 1, hjust=1, size=8))

^{由reprex package (v2.0.1) 于 2021 年 12 月 21 日创建}

【讨论】：

谢谢。 ggbeeswarm 帮助了我！但是，在您的解决方案中，例如在 caud 中有一些蓝点，然后是红点，然后是蓝点，这有点混乱。我设置 cex=0.7，dodge.width=0.1。这改善了情节，但由于某种原因，在某些情况下红点出现在蓝色之上，而在某些情况下则相反（即使我将其设为有序因子）。为了解决这个问题，我必须创建一个空的 ggplot 对象，然后使用 "data = df %>% filter(Shielding=="XRB") 等绘制三个单独的蜂群。结果图正是我想要的！谢谢!
不确定是否有必要包含多个图，更新了答案以显示如何使用 geom_beeswarm ，dodge.width 参数实现这一点。
这确实是一个更优雅的解决方案。但是，如果您使用 geom_jitter 查看我的示例和您的答案：在 LAO90 中，蓝点在红点前面，而在下一组 RAO 中，蓝点在红点前面，然后蓝色再次在 RAO-CAUD 中的红色前面。我觉得他们交替哪个颜色组在前面有点奇怪。像你一样设置 alpha=0.5 使它不那么明显，但它仍然不一致的是顶层中的颜色。
这可以通过创建绘图顺序变量来实现：请参阅更新的答案。
是的！它可以工作，实际上它甚至可以在不添加“秩序”美学的情况下工作。显然ggplot根据行号堆叠点（我不知道），你使用“排列”的解决方案很好地解决了这个问题。再次感谢您的回答和更新。 C

【解决方案2】：

您可以在将 x 转换为连续变量后手动计算偏移量。

library(tidyverse)

## prepare x
df <- df %>%
  group_by(projection) %>%
  mutate(
    group = cur_group_id(),
    x = rep(seq(unique(group) - .1, unique(group) + .1, 0.05), 3)
  ) 

ggplot(df, aes(x = x, y = E.DAP, color = Shielding)) +
  ## you can use geom_point now
  geom_point(alpha = 0.8) +
  theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust = 1, size = 8)) +
  labs(title = "mSv/DAP according to projection",
    ## use x = NULL and color = NULL instead of your theme element
    y = "mSv/mGy*cm2 ", x = NULL, color = NULL
  ) +
  theme(legend.position = c(0.85, 0.8), legend.key = element_rect(colour = "transparent", fill = "transparent")) +
  scale_color_manual(values = c(c("#69ac5c", "#5b86c3", "#c7533b"))) +
  ## you will now need to set the breaks and labels 
  scale_x_continuous(breaks = seq_len(max(df$group)), labels = sort(unique(df$projection)))

【讨论】：