使用ggplot2的facet_grid分面时出现行缺失问题的解决求助

需要绘制20名参与者（分为a、b两组各10人）40天内三类事件（missing、no、yes）的数量分布。未分面时绘图显示正常，但使用facet_grid按group分面后出现行缺失情况。

示例代码：

library(dplyr)
library(tidyr)
library(ggplot2)

set.seed(1234)

d <- data.frame(id = NULL,
                present = NULL)

# create data. 20 ids with forty obs each
for(i in 0:19) {
  d[i*40+(1:40), "present"] <- sample(x = c("no","yes","missing"),
                                      size = 40,
                                      prob = c(0.25, 0.4, 0.35),
                                      replace = T)
  d[i*40+(1:40), "id"] <- rep(paste0("id0",1+i),40)
}

# add group
d$group <- factor(rep(letters[1:2],each=400))

# now get summaries of proportions of each of the three variables, then uncount to recreate the individual rows but this time sorted 
d %>%
  group_by(id, group, present) %>%
    count  %>%
      uncount(weights = n) %>%
        ungroup %>%
          tibble::add_column(obsNum = rep(1:40,times = 20)) -> d

# now rank
d %>%
  group_by(id) %>%
    add_count(wt = present == "missing", # this counts the number of missing for each person and adds that count as a column
              name = "miss_n") %>%
      add_count(wt = present == "no", # does the same thing as above but for number of no values
                name = "no_n") %>%
        ungroup %>%
          arrange(-miss_n, -no_n, present) %>%
            mutate(id = factor(id) %>% fct_inorder()) %>%
              mutate(obsNum = row_number(), # adds a new set of row numbers within each id numbers. 
                     .by = id) -> d_ranked

未分面绘图代码（效果正常）：

d_ranked %>%
  ggplot(mapping = aes(x = obsNum, 
                       y = id, 
                       colour = present)) +
         geom_point(shape = 15) +
         scale_colour_manual(values = c("yellow", "red", "black")) +
         labs(x = "observation number")

分面后出现行缺失的代码：

d_ranked %>%
  ggplot(mapping = aes(x = obsNum, 
                       y = id, 
                       colour = present)) +
        geom_point(shape = 15) +
        facet_grid(.~group) +
        scale_colour_manual(values = c("yellow", "red", "black")) +
        labs(x = "observation number")

你的id因子是全局排序后生成的（按所有参与者的miss_n和no_n排序），导致a、b两组的id在因子水平中交错分布。当使用facet_grid分面后，每个面板仅包含对应组的id数据，但y轴仍会显示所有20个id的因子水平，不属于当前组的id位置就会呈现“空行”，看起来像是行缺失。

方式1：按组内排序生成id因子 #

修改数据处理的排序步骤，在group分组内对id进行排序，这样每个组的id在因子水平中是连续的，分面后每个面板的y轴只会显示当前组的id（无空行）：

d %>%
  group_by(id, group, present) %>%
  count %>%
  uncount(weights = n) %>%
  ungroup %>%
  tibble::add_column(obsNum = rep(1:40,times = 20)) -> d

# 修改排序逻辑：按group分组后，在组内排序id
d %>%
  group_by(id) %>%
  add_count(wt = present == "missing", name = "miss_n") %>%
  add_count(wt = present == "no", name = "no_n") %>%
  ungroup %>%
  # 先按group分组，再组内按-miss_n、-no_n排序
  arrange(group, -miss_n, -no_n, present) %>%
  # 按排序后的顺序生成因子，组内id连续
  mutate(id = factor(id) %>% fct_inorder()) %>%
  mutate(obsNum = row_number(), .by = id) -> d_ranked

使用原分面代码绘图后，每个面板会显示对应组的10个连续id，无空行。

方式2：分面时适配y轴显示 #

如果你希望每个面板仅显示当前组的id且无空行，也可以不修改数据，直接在分面时设置scales = "free_y"和space = "free_y"：

d_ranked %>%
  ggplot(mapping = aes(x = obsNum, 
                       y = id, 
                       colour = present)) +
  geom_point(shape = 15) +
  facet_grid(.~group, scales = "free_y", space = "free_y") +
  scale_colour_manual(values = c("yellow", "red", "black")) +
  labs(x = "observation number")

scales = "free_y"让每个面板的y轴只显示有数据的id，space = "free_y"让每个面板的y轴高度适配id数量，避免空行占用空间。

内容的提问来源于stack exchange，提问作者llewmills