如何在使用对数刻度时绘制 ggplot 的 geom_errorbar

Question

我想在线性和对数y轴上绘制带有标准差误差线的摘要线图。 在线性尺度上，我可以很容易地做到这一点，但是在生成有效的标准差误差线时，如何考虑对数尺度呢？ 也许我需要手动计算几何标准差？不确定如何实现这一点。

虚拟数据：

library(dplyr)
dat <- structure(list(time = c(0, 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 
                             5, 5.5, 6, 6.5, 7, 7.5, 8, 8.5, 9, 9.5, 10, 10.5, 11, 11.5, 12, 
                             0, 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6, 6.5, 7, 7.5, 
                             8, 8.5, 9, 9.5, 10, 10.5, 11, 11.5, 12, 0, 0.5, 1, 1.5, 2, 2.5, 
                             3, 3.5, 4, 4.5, 5, 5.5, 6, 6.5, 7, 7.5, 8, 8.5, 9, 9.5, 10, 10.5, 
                             11, 11.5, 12, 0, 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 
                             6, 6.5, 7, 7.5, 8, 8.5, 9, 9.5, 10, 10.5, 11, 11.5, 12, 0, 0.5, 
                             1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6, 6.5, 7, 7.5, 8, 8.5, 
                             9, 9.5, 10, 10.5, 11, 11.5, 12, 0, 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 
                             4, 4.5, 5, 5.5, 6, 6.5, 7, 7.5, 8, 8.5, 9, 9.5, 10, 10.5, 11, 
                             11.5, 12), dv = c(0, 0.18432616, 0.88526902, 1.2296906, 1.28501176, 
                                               1.2139172, 1.09455226, 0.96353803, 0.83722931, 0.72220134, 0.62040715, 
                                               0.53169995, 0.45505514, 0.38915199, 0.33264156, 0.2842622, 0.24288196, 
                                               0.20750707, 0.1772753, 0.15144348, 0.12937353, 0.11051874, 0.09441129, 
                                               0.08065113, 0.06889633, 0, 0.18646195, 0.9048901, 1.28471715, 
                                               1.37742791, 1.33879388, 1.24474175, 1.13178462, 1.01703554, 0.90810807, 
                                               0.80800831, 0.71754717, 0.63652547, 0.56431222, 0.50012324, 0.44315229, 
                                               0.39262987, 0.34784693, 0.30816176, 0.27299918, 0.24184632, 0.21424719, 
                                               0.18979702, 0.16813683, 0.14894842, 0, 0.18294322, 0.87292448, 
                                               1.19567138, 1.22915435, 1.14027058, 1.00823429, 0.8693954, 0.73934779, 
                                               0.62380257, 0.52390843, 0.43883218, 0.36699104, 0.30662467, 0.25604644, 
                                               0.21374115, 0.1783911, 0.14887033, 0.12422626, 0.10365757, 0.08649245, 
                                               0.07216874, 0.06021662, 0.05024367, 0.0419223, 0, 0.18697047, 
                                               0.90966337, 1.29828203, 1.40059765, 1.37067756, 1.2838258, 1.17643544, 
                                               1.06571995, 0.95947803, 0.86092175, 0.77106034, 0.68987364, 0.61688687, 
                                               0.55144953, 0.49286823, 0.44046784, 0.39361759, 0.35174019, 0.31431303, 
                                               0.28086579, 0.25097654, 0.22426743, 0.20040041, 0.17907321, 0, 
                                               0.18531778, 0.89429381, 1.25485411, 1.32695875, 1.27013735, 1.16158786, 
                                               1.03796132, 0.91603064, 0.80290267, 0.7010534, 0.61080177, 0.53151718, 
                                               0.46220211, 0.40176719, 0.34915557, 0.30339446, 0.26361157, 0.22903568, 
                                               0.19899009, 0.17288363, 0.15020104, 0.13049386, 0.11337208, 0.09849667, 
                                               0, 0.18383905, 0.88088906, 1.21756741, 1.26499344, 1.18736204, 
                                               1.0632291, 0.92915041, 0.80123546, 0.68577026, 0.58443313, 0.49683915, 
                                               0.42176711, 0.35773895, 0.30328282, 0.25704292, 0.21781669, 0.18455866, 
                                               0.15636983, 0.13248206, 0.11224132, 0.09509189, 0.0805622, 0.06825233, 
                                               0.05782327)), class = "data.frame", row.names = c(NA, -150L))

# Summarise
dat_summarised <- dat |> 
  group_by(time) |> 
  summarise(mean = mean(dv, na.rm = T), 
            sd = sd(dv, na.rm = T))

# Plot linear (this works)
dat_summarised |> 
    ggplot(aes(x = time, y = mean)) +
    geom_point(size = 3) +
    geom_line(linewidth = 0.8) +
    geom_errorbar(aes(ymax = mean + sd, ymin = mean - sd), width = 0.1) +
    theme_classic(base_size = 20)

# Plot log (this doesn't work)
dat_summarised |> 
  ggplot(aes(x = time, y = mean)) +
  geom_point(size = 3) +
  geom_line(linewidth = 0.8) +
  geom_errorbar(aes(ymax = mean + sd, ymin = mean - sd), width = 0.1) +
  scale_y_log10() +
  theme_classic(base_size = 20)

- LucaS

---

是的，我忘记了那些零。但那不是我的主要关注点。我的主要关注点是标准差是在线性而不是对数尺度上计算的。也许我只是把自己搞糊涂了，因为只有绘图尺度在改变，而实际值并没有改变。如果是这样的话，那么原始代码实际上不需要做任何改变。 - undefined

我还遇到了这个回答，但对我来说并没有澄清问题：https://stackoverflow.com/questions/65525087/errorbars-in-a-log-scale-plot - undefined

对数变换是单调的。在我的答案中，变换log(dv + 1L)是在使用geom_errorbar()之前完成的，因此该函数使用了sd和mean的变换值。你应该关注零值。 - undefined

Answer 1

在前一步中，将mean和sd变量进行转换如何？

suppressPackageStartupMessages({
  library(dplyr)
  library(ggplot2)
})

dat |> 
  summarise(mean = mean(log(dv + 1L), na.rm = TRUE), 
            sd = sd(log(dv + 1L), na.rm = TRUE), 
            .by = time) |>
  ggplot(aes(x = time, y = mean)) +
  geom_point(size = 3L) +
  geom_line(linewidth = .8) +
  geom_errorbar(aes(ymax = mean + sd, ymin = mean - sd), width = .1) +
  # scale_y_log10() +
  theme_classic(base_size = 20L)

^{创建于2023-10-11，使用reprex v2.0.2。}

由于一些零值，您的对数转换会导致问题：log(0)为-Inf。您可以尝试mean(log(dv + 1L), na.rm = TRUE)和sd = sd(log(dv + 1L), na.rm = TRUE)。

- Friede

Answer 2

你的问题不是误差条，而是你的均值中有一个0。0不能转换为对数坐标图中的位置，因为它会变成负无穷。

如果你删除第一个数据点，或者用一个小值替换它，应该就可以解决问题了。

- backboned

Answer 3

或者，您可以使用函数log(x - min(x) - 1)在scale_y_continuous中消除对数中的零和负数，就像这样：

library(dplyr)
library(ggplot2)

dat_summarised |> 
  ggplot(aes(x = time, y = mean)) +
  geom_point(size = 3) +
  geom_line(linewidth = 0.8) +
  geom_errorbar(aes(ymax = mean + sd, ymin = mean - sd), width = 0.1) +
  scale_y_continuous(breaks = \(x) log(x - (min(x) - 1))) +
  theme_classic(base_size = 20)

^{2023-10-11创建，使用reprex v2.0.2生成}

- Quinten