为数据集添加含新增唯一分类变量的排列组合

我有一个包含多分类变量和数值变量的数据集，分类变量存在若干唯一组合，但并非所有排列组合都存在（无对应值的组合无需保留）。需要为所有分类变量添加名为[no selection]的水平，生成包含以下两类组合的新数据集：

1. 原数据中的所有有效组合
2. 所有仅将某一个分类变量替换为[no selection]的唯一组合（其余分类变量保持原有效组合的取值）
   新组合中的数值变量值设为0（空值后续可自行替换）。

提供的R数据集结构如下：

structure(list(`Primary Outcome` = structure(c(1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L), levels = c("Laboratory-Confirmed Influenza", 
"Influenza-Like Illness", "Complications from Influenza"), class = "factor"), 
    `Virus (Sub)Type` = structure(c(1L, 1L, 2L, 2L, 3L, 3L, 4L, 
    4L, 5L, 5L, 6L), levels = c("Influenza A and B", "Influenza A", 
    "Influenza B", "A(H1N1)pdm09", "A(H3N2)", "B/Yamagata", "B/Victoria", 
    "B/Wisconsin"), class = "factor"), Season = structure(c(1L, 
    1L, 1L, 1L, 1L, 3L, 3L, 3L, 3L, 3L, 3L), levels = c("All Seasons", 
    "2019/20", "2011/12", "2014/15", "2018/19", "2017/18", "2013/14", 
    "2012/13", "2015/16", "2016/17", "2012", "2013", "2018", 
    "2014", "2019", "2016", "2017", "2010-2019", "2015"), class = "factor"), 
    `Meta-Analysis Model` = structure(c(1L, 2L, 1L, 2L, 1L, 2L, 
    1L, 2L, 1L, 2L, 1L), levels = c("Fixed Effect Model", "Random Effects Model"
    ), class = "factor"), k = c(398, 398, 126, 126, 212, 7, 2, 
    2, 10, 10, 2), Q = c(1236.50029339945, 1236.50029339945, 
    352.741540498925, 352.741540498925, 380.231921977758, 11.3144127452121, 
    2.68421098744275, 2.68421098744275, 5.75656937709005, 5.75656937709005, 
    1.74656825505281), p = c(5.89614495232528e-87, 5.89614495232528e-87, 
    1.38614414562279e-23, 1.38614414562279e-23, 8.10697738675785e-12, 
    0.0791317973899664, 0.101347413621071, 0.101347413621071, 
    0.764012962020228, 0.764012962020228, 0.186308732770133), 
    `Vaccine Matching Status` = c("No Stratification", "No Stratification", 
    "No Stratification", "No Stratification", "No Stratification", 
    "No Stratification", "No Stratification", "No Stratification", 
    "No Stratification", "No Stratification", "No Stratification"
    ), `Risk Group` = c("General Population (Aged 5-64)", "General Population (Aged 5-64)", 
    "General Population (Aged 5-64)", "General Population (Aged 5-64)", 
    "General Population (Aged 5-64)", "General Population (Aged 5-64)", 
    "General Population (Aged 5-64)", "General Population (Aged 5-64)", 
    "General Population (Aged 5-64)", "General Population (Aged 5-64)", 
    "General Population (Aged 5-64)")), row.names = c(NA, -11L
), class = c("tbl_df", "tbl", "data.frame"))

注意：expand.grid无法满足需求，因为不需要现有变量的全排列，仅需保留原有效组合及含[no selection]的组合。

核心思路是针对每个分类变量生成单独替换为[no selection]的组合，再与原数据合并去重，具体步骤如下：

步骤1：定义变量类型 #

先明确分类变量和数值变量的范围：

# 加载数据集
df <- structure(...) # 替换为你提供的数据集结构

# 分类变量列名
cat_vars <- c("Primary Outcome", "Virus (Sub)Type", "Season", "Meta-Analysis Model", 
              "Vaccine Matching Status", "Risk Group")
# 数值变量列名
num_vars <- c("k", "Q", "p")

步骤2：生成含新水平的组合 #

循环每个分类变量，生成仅该变量为[no selection]的新组合，数值变量设为0：

# 初始化结果列表，先加入原数据
result_list <- list(df)

# 遍历每个分类变量生成新组合
for (var in cat_vars) {
  temp_df <- df
  # 替换当前分类变量为新水平
  temp_df[[var]] <- "[no selection]"
  # 数值变量设为0
  temp_df[num_vars] <- 0
  # 加入结果列表
  result_list <- c(result_list, list(temp_df))
}

步骤3：合并去重 #

将所有组合合并，去除重复行：

# 合并所有数据框
final_df <- dplyr::bind_rows(result_list)

# 去重，保留唯一组合
final_df <- dplyr::distinct(final_df)

# 可选：将分类变量重新转为因子（包含原水平+新水平）
for (var in cat_vars) {
  final_df[[var]] <- factor(final_df[[var]], 
                            levels = c(levels(df[[var]]), "[no selection]"))
}

代码说明 #

• 保留原数据的所有有效组合
• 对每个分类变量生成一组新组合：仅该变量取值为[no selection]，其余分类变量保持原组合取值，数值变量设为0
• 合并后去重，避免不同变量替换后出现重复组合
• 可选步骤将分类变量转为包含新水平的因子，保证因子类型的一致性

内容的提问来源于stack exchange，提问作者Hunter