向R函数传递参数失败：变量与列范围传递报错

问题背景 #

需要将一段处理缺失值的R代码封装为函数，传入参数包括：数据框名、ResponseId变量名、一组连续变量范围（如PMS_1:PMS_45）。但函数调用时提示PMS_1不存在，参数传递方式有误。

可运行的原始代码 #

set.seed(1024)

premice_1 <- subset(emu_raw, select=c(ResponseId, PMS_1:PMS_45))
premice_2 <- mice(premice_1, maxit=0, pri=F)
pred <- premice_2$pred
pred[,c("ResponseId")] <- 0
meth <- premice_2$meth
meth[c("ResponseId")] <- ""

postmice_a <- mice(premice_1, m=1, maxit=10, printFlag=TRUE, pred=pred, 
                   meth=meth, seed=4201)
postmice_b <- mice::complete(postmice_a, "long", include=FALSE)
postmice_b$.imp <- NULL
postmice_b$.id <- NULL

emu_raw <- merge(emu_raw, postmice_b)

出错的函数及调用 #

调用方式：missing_values(emu_raw, ResponseId, PMS_1:PMS_45)，错误提示object 'PMS_1' not found。

出错函数：

missing_values <- function(df, ResponseId, measure_range) {

set.seed(1024)

premice_1 <- subset(df, select=c(ResponseId, measure_range))
premice_2 <- mice(premice_1, maxit=0, pri=F)
pred <- premice_2$pred
pred[,c("ResponseId")] <- 0
meth <- premice_2$meth
meth[c("ResponseId")] <- ""

postmice_a <- mice(premice_1, m=1, maxit=10, printFlag=TRUE, pred=pred, 
                   meth=meth, seed=4201)
postmice_b <- mice::complete(postmice_a, "long", include=FALSE)
postmice_b$.imp <- NULL
postmice_b$.id <-NULL

return(postmice_b)
}

尝试用方括号替代subset()后仍报错：

missing_values <- function(df, ResponseId, measure_range) {

set.seed(1024)

premice_1 <- df[,c(ResponseId, measure_range)]
premice_2 <- mice(premice_1, maxit=0, pri=F)
pred <- premice_2$pred
pred[,c(ResponseId)] <- 0
meth <- premice_2$meth
meth[c("ResponseId")] <- ""

postmice_a <- mice(premice_1, m=1, maxit=10, printFlag=TRUE, pred=pred, 
                   meth=meth, seed=4201)
postmice_b <- mice::complete(postmice_a, "long", include=FALSE)
postmice_b$.imp <- NULL
postmice_b$.id <- NULL

return(postmice_b)
}

错误原因 #

直接在函数调用中传入PMS_1:PMS_45时，R会尝试在全局环境中查找名为PMS_1的对象，而非将其解析为数据框内的列名范围。同时，ResponseId的传递也未处理非标准求值的问题，导致列选择逻辑失效。

解决方案 #

提供两种可行的修正方式，按需选择：

方案一：使用非标准求值（推荐，保持调用直观性） #

借助rlang包的非标准求值工具，让R在数据框的环境中解析变量名和范围：

library(mice)
library(rlang)

missing_values <- function(df, ResponseId, measure_range) {
  set.seed(1024)
  
  # 用{{}}注入变量，解析数据框内的列名和范围
  premice_1 <- df |>
    dplyr::select({{ ResponseId }}, {{ measure_range }})
  
  premice_2 <- mice(premice_1, maxit=0, pri=F)
  pred <- premice_2$pred
  
  # 将ResponseId转为字符串，处理pred和meth矩阵
  resp_id_str <- as_name(enquo(ResponseId))
  pred[, resp_id_str] <- 0
  meth <- premice_2$meth
  meth[resp_id_str] <- ""
  
  postmice_a <- mice(premice_1, m=1, maxit=10, printFlag=TRUE, pred=pred, 
                     meth=meth, seed=4201)
  postmice_b <- mice::complete(postmice_a, "long", include=FALSE)
  postmice_b$.imp <- NULL
  postmice_b$.id <- NULL
  
  return(postmice_b)
}

调用方式（无需引号，和原始代码写法一致）：

# 调用函数
imputed_data <- missing_values(emu_raw, ResponseId, PMS_1:PMS_45)
# 合并回原数据框
emu_raw <- merge(emu_raw, imputed_data)

方案二：使用字符串传递列名范围 #

若不想依赖rlang，可直接生成列名字符串向量传递：

library(mice)

missing_values <- function(df, ResponseId, measure_range) {
  set.seed(1024)
  
  # 合并列名字符串
  cols <- c(ResponseId, measure_range)
  premice_1 <- df[, cols, drop = FALSE]
  
  premice_2 <- mice(premice_1, maxit=0, pri=F)
  pred <- premice_2$pred
  pred[, ResponseId] <- 0
  meth <- premice_2$meth
  meth[ResponseId] <- ""
  
  postmice_a <- mice(premice_1, m=1, maxit=10, printFlag=TRUE, pred=pred, 
                     meth=meth, seed=4201)
  postmice_b <- mice::complete(postmice_a, "long", include=FALSE)
  postmice_b$.imp <- NULL
  postmice_b$.id <- NULL
  
  return(postmice_b)
}

调用方式：

# 生成列名字符串向量
measure_cols <- paste0("PMS_", 1:45)
# 调用函数
imputed_data <- missing_values(emu_raw, "ResponseId", measure_cols)
# 合并回原数据框
emu_raw <- merge(emu_raw, imputed_data)

内容的提问来源于stack exchange，提问作者JRDubbleu