如何在R中基于最大长度为数据框补全缺失行

问题场景 #

实验数据中，参与者逐词阅读句子并回答YES/NO，一旦回答NO，句子立即结束，后续词的记录缺失。需要补全从NO对应的word_position到max_length之间的所有缺失行，保留participant、sentence_id等静态信息，新行的response设为NO。

解决方案（使用tidyverse工具） #

首先加载所需包：

library(tidyverse)

对示例数据进行处理：

test_filled <- test %>%
  # 按「参与者+句子」分组，确保每组独立处理
  group_by(participant, sentence_id) %>%
  # 计算每组中第一个NO出现的位置，以及句子的最大长度
  mutate(
    first_no_pos = ifelse(any(response == "NO"), min(word_position[response == "NO"]), NA),
    group_max = max(max_length)
  ) %>%
  # 补全该句子所有应存在的word_position行
  complete(word_position = 1:group_max) %>%
  # 填充参与者、句子ID等组内固定信息
  fill(participant, sentence_id, group_max, first_no_pos, .direction = "downup") %>%
  # 设置响应值：原有响应保留，NO位置及之后的新增行设为NO
  mutate(
    response = case_when(
      !is.na(response) ~ response,
      word_position >= first_no_pos ~ "NO",
      TRUE ~ NA_character_
    ),
    max_length = group_max
  ) %>%
  # 清理临时辅助列，取消分组
  select(-first_no_pos, -group_max) %>%
  ungroup()

补充：填充缺失的word列 #

如果需要将缺失的word列也补全为句子的正确词汇，可以先建立句子-词位的映射表，再关联到数据中：

# 构建句子完整词序列的映射表
word_lookup <- tribble(
  ~sentence_id,        ~word_position, ~word,
  "dog_sentence",      1,              "the",
  "dog_sentence",      2,              "dog",
  "dog_sentence",      3,              "went",
  "dog_sentence",      4,              "home.",
  "plant_sentence",    1,              "I",
  "plant_sentence",    2,              "watered",
  "plant_sentence",    3,              "my",
  "plant_sentence",    4,              "plants",
  "plant_sentence",    5,              "today."
)

# 关联映射表补全word列
test_filled_with_words <- test_filled %>%
  left_join(word_lookup, by = c("sentence_id", "word_position")) %>%
  mutate(word = coalesce(word.x, word.y)) %>%
  select(-word.x, -word.y)

效果验证 #

查看参与者001的plant_sentence补全结果：

test_filled_with_words %>% filter(participant == "001", sentence_id == "plant_sentence")

输出结果：

内容的提问来源于stack exchange，提问作者microcastle