R: t.test 在 facet_grid (ggplot)

Question

這是一個非常具體的問題，但我已經擁有並使用了這個詳細且運行良好的代碼，所以我希望找到調整它所需的微小更改，並使其適用於下一個復雜級別。 我得到了什么：

library(ggplot2)
library(ggpubr)
head(ToothGrowth)
ToothGrowth$dose <- as.factor(ToothGrowth$dose)
# add a grouping ID for measured individuals:
ToothGrowth$ID <- rep(c(1:30),2)

# The code I am using now (basically a solution I got from my former question answered by Allan Cameron (user:12500315)):
ggplot(ToothGrowth, aes(supp, len, fill = dose, alpha = supp)) +
  geom_boxplot() +
  scale_fill_manual(name   = "Dosis", 
                    labels = c("0.5", "1", "2"), 
                    values = c("darkorange2", "olivedrab", "cadetblue4")) +
  scale_alpha_discrete(range = c(0.5, 1), 
                       guide = guide_none()) +
  geom_line(inherit.aes = FALSE, 
            aes(supp, len, group = ID), 
            color = "gray75") +
  geom_text(data = data.frame(
    x    = 1.5, 
    y    = 40, 
    dose = c("0.5", "1", "2"),
    pval = sapply(c("0.5", "1", "2"), function(x) {
      round(t.test(len ~ supp, 
                   data = ToothGrowth[ToothGrowth$dose == x,],
                   paired = TRUE)$p.val, 4)})), 
    inherit.aes = FALSE,
    aes(x = 1.5, y = 40, label = paste("T test: p value =", pval)), 
    check_overlap = TRUE) +
  facet_grid(~dose) +
  theme_classic() +
  theme(legend.position = "top",
        strip.background = element_rect(fill = "gray95", size = 0.25))

# Follow-up question:
# What I want to do next: having another facetting variable ('researcher')
ToothGrowth_1 <- ToothGrowth
# create a random numerical factor to multiply measures with and then enlarge the dataset by a second set of measurements from a different 'researcher':
r <- runif(60, min=0, max=3)
ToothGrowth_1$len <- ToothGrowth_1$len*r
ToothGrowth$researcher <- "A"
ToothGrowth_1$researcher <- "B"
ToothGrowth_total <- rbind(ToothGrowth, ToothGrowth_1)

現在，我想 plot 與上面相同的 plot ，但對兩個“研究人員”組（A 與 B）進行水平切面拆分。 我通過創建和交互術語“researcher”和“dose”並用 facet_wrap 替換 facet_grid 找到了一個解決方法，但我更願意看到 facet_grid 的解決方案，因為它使其他一切變得更容易。 感謝您的幫助，非常感謝！

Answer 1

感謝您發布后續。

做到這一點的自然方法是map兩個級別，盡管我認為與其完全重寫來實現這一點，我可能只會使用 2 個sapply調用 - 一個用於新因素的每個級別：

ggplot(ToothGrowth_total, aes(supp, len, fill = dose, alpha = supp)) +
  geom_boxplot() +
  scale_fill_manual(name   = "Dosis", 
                    labels = c("0.5", "1", "2"), 
                    values = c("darkorange2", "olivedrab", "cadetblue4")) +
  scale_alpha_discrete(range = c(0.5, 1), 
                       guide = guide_none()) +
  geom_line(inherit.aes = FALSE, 
            aes(supp, len, group = ID), 
            color = "gray75") +
  geom_text(data = data.frame(
    x    = 1.5, 
    y    = c(40, 40, 40, 70, 70, 70), 
    researcher = c("A", "A", "A", "B", "B", "B"),
    dose = c("0.5", "1", "2", "0.5", "1", "2"),
    pval = c(sapply(c("0.5", "1", "2"), function(x) {
            round(t.test(len ~ supp, 
            data = subset(ToothGrowth_total, dose == x & researcher == "A"),
            paired = TRUE)$p.val, 4)}), 
            sapply(c("0.5", "1", "2"), function(x) {
            round(t.test(len ~ supp, 
            data = subset(ToothGrowth_total, dose == x & researcher == "B"),
            paired = TRUE)$p.val, 4)}))),
    inherit.aes = FALSE,
    aes(x = x, y = y, label = paste("T test: p value =", pval)), 
    check_overlap = TRUE) +
  facet_grid(researcher~dose, scales = "free_y") +
  theme_classic() +
  theme(legend.position = "top",
        strip.background = element_rect(fill = "gray95", size = 0.25))

Answer 2

如果我沒記錯的話，我實際上找到了一種更簡單的方法：

ToothGrowth_total$researcher_dose <- interaction(ToothGrowth_total$researcher, ToothGrowth_total$dose)

ggplot(ToothGrowth_total, aes(supp, len, fill = dose, alpha = supp)) +
  geom_boxplot() +
  scale_fill_manual(name   = "Dosis", 
                    labels = c("0.5", "1", "2"), 
                    values = c("darkorange2", "olivedrab", "cadetblue4")) +
  scale_alpha_discrete(range = c(0.5, 1), 
                       guide = guide_none()) +
  geom_line(inherit.aes = FALSE, 
            aes(supp, len, group = ID), 
            color = "gray75") +
  # geom_text(data = data.frame(
  #   x    = 1.5, 
  #   y    = c(40, 40, 40, 70, 70, 70), 
  #   researcher = c("A", "A", "A", "B", "B", "B"),
  #   dose = c("0.5", "1", "2", "0.5", "1", "2"),
  #   pval = c(sapply(c("0.5", "1", "2"), function(x) {
  #     round(t.test(len ~ supp, 
  #                  data = subset(ToothGrowth_total, dose == x & researcher == "A"),
  #                  paired = TRUE)$p.val, 4)}), 
  #     sapply(c("0.5", "1", "2"), function(x) {
  #       round(t.test(len ~ supp, 
  #                    data = subset(ToothGrowth_total, dose == x & researcher == "B"),
  #                    paired = TRUE)$p.val, 4)}))),
  #   inherit.aes = FALSE,
  #   aes(x = x, y = y, label = paste("T test: p value =", pval)), 
  #   check_overlap = TRUE) +
# => instead subsituted by: 
  stat_compare_means(aes(x="researcher_dose"), method = "t.test", paired = TRUE)+ 
  facet_grid(researcher~dose, scales = "free_y") +
  theme_classic() +
  theme(legend.position = "top",
        strip.background = element_rect(fill = "gray95", size = 0.25))

我希望我在這里沒有遺漏任何重要的東西，但它會產生相同的t.test結果，因此我認為它是正確的。 如果沒有，請告訴我。 唯一的區別是“researcher_dose”現在也顯示為 x 軸標簽。