使用tbl_summary生成2×2表后，如何通过卡方检验获取正确P值？

你的需求是对分组求和后的2×2表做卡方检验并得到单个P值，当前add_p()会为每行生成P值，原因是tbl_summary()将correct_answers和incorrect_answers视为两个独立变量，而非组成列联表的两个类别。以下是两种可行的解决方法：

方法一：保留原汇总表样式，添加卡方P值 #

先手动计算分组求和后的列联表卡方检验结果，再将P值添加到原汇总表底部：

library(gtsummary)
library(tidyverse)

test <- data.frame("With_assistant" = c(TRUE, TRUE, TRUE, TRUE, TRUE, FALSE, FALSE, FALSE, FALSE, TRUE, FALSE),
                   "correct_answers" = c(2,4,5,6,1,2,7,2,1,2,3),
                   "incorrect_answers" = c(1,2,1,5,3,1,2,5,3,2,4))

# 1. 计算分组求和的2×2列联表，并做卡方检验
cross_tab <- test %>%
  group_by(With_assistant) %>%
  summarize(correct = sum(correct_answers), incorrect = sum(incorrect_answers))
chisq_result <- chisq.test(as.matrix(cross_tab[, -1]))

# 2. 生成原汇总表并添加卡方P值
test %>%
  tbl_summary(
    by = With_assistant,
    type = list(c(correct_answers, incorrect_answers) ~ "continuous"),
    statistic = list(c(correct_answers, incorrect_answers) ~ "{sum}") 
  ) %>%
  # 在表底添加卡方检验P值
  add_stat(
    fns = ~ ifelse(.variable %in% c("correct_answers", "incorrect_answers"),
                   paste("卡方检验P值:", round(chisq_result$p.value, 3)),
                   NA),
    location = "bottom"
  ) %>%
  # 清理多余行
  modify_table_body(~ .x %>% filter(!(row_type == "statistic" & is.na(label))) %>% distinct())

方法二：用tbl_cross()直接生成带卡方检验的列联表 #

将数据转换为适合列联表的长格式，使用tbl_cross()直接生成符合卡方检验要求的表格：

library(gtsummary)
library(tidyverse)

test <- data.frame("With_assistant" = c(TRUE, TRUE, TRUE, TRUE, TRUE, FALSE, FALSE, FALSE, FALSE, TRUE, FALSE),
                   "correct_answers" = c(2,4,5,6,1,2,7,2,1,2,3),
                   "incorrect_answers" = c(1,2,1,5,3,1,2,5,3,2,4))

# 将数据展开为每个答案条目单独一行的格式
test_long <- test %>%
  mutate(id = row_number()) %>%
  pivot_longer(cols = c(correct_answers, incorrect_answers),
               names_to = "answer_type", values_to = "count") %>%
  uncount(count) %>%
  mutate(answer_type = ifelse(answer_type == "correct_answers", "正确答案", "错误答案"))

# 生成列联表并添加卡方检验P值
test_long %>%
  tbl_cross(row = answer_type, col = With_assistant) %>%
  add_p(test = ~ chisq.test(.))

两种方法的卡方检验结果一致，你可以根据需要选择表格样式。

内容的提问来源于stack exchange，提问作者MegStats