变量高效重编码方法、均值按比例计算及批量同规则重编码问询

Hey there! Let's tackle your two data analysis questions using R, since your dataset is structured in R's format. Here's a step-by-step breakdown:

高效变量重编码方法 #

Below are several practical, efficient methods for recoding variables in R, sorted by flexibility and readability:

• dplyr::case_when()：The most versatile option for multi-condition recoding, with clean, human-readable syntax that performs well even on large datasets.
  Example (convert 1→0, 2→1, keep NA values):

  library(dplyr)
  df <- df %>% mutate(var_recoded = case_when(
    var == 1 ~ 0,
    var == 2 ~ 1,
    TRUE ~ NA_real_
  ))
  

• Base R ifelse()：Perfect for simple binary recoding, with concise code:

  df$var_recoded <- ifelse(df$var == 1, 0, ifelse(df$var == 2, 1, NA))
  

• car::recode()：A compact tool for quick categorical recoding:

  library(car)
  df$var_recoded <- recode(df$var, "1=0; 2=1; else=NA")
  

按比例计算均值 #

This typically falls into two common scenarios:

Scenario 1: Weighted Mean (calculate mean using proportional weights) #

If you have a weight column (e.g., weight), compute the weighted mean of your target variable:

# Base R
weighted_mean <- weighted.mean(df$var, df$weight, na.rm = TRUE)

# dplyr approach
df %>% summarize(weighted_mean = weighted.mean(var, weight, na.rm = TRUE))

Scenario 2: Grouped Mean with Proportion Distribution #

Calculate the mean per group, along with each group's proportion of the total sample:

df %>%
  group_by(group) %>%
  summarize(
    var_mean = mean(var, na.rm = TRUE),
    group_proportion = n() / nrow(df)
  )

Your dataset has variables named bhs1_1 to bhs1_10 (all with values 1, 2, or NA). Using dplyr::across() is the most efficient way to recode all these variables at once, without repeating code for each column.

Example: Recode 1→0, 2→1 (with new columns for recoded values) #

library(dplyr)

df_bhs1 <- df_bhs1 %>%
  mutate(across(starts_with("bhs1_"), ~ case_when(
    .x == 1 ~ 0,
    .x == 2 ~ 1,
    TRUE ~ NA_real_
  ), .names = "{.col}_recoded")) # Adds "_recoded" suffix to new columns

If you want to overwrite the original variables directly: #

df_bhs1 <- df_bhs1 %>%
  mutate(across(starts_with("bhs1_"), ~ case_when(
    .x == 1 ~ 0,
    .x == 2 ~ 1,
    TRUE ~ NA_real_
  )))

Alternative: Recode to categorical labels (e.g., "Agree"/"Disagree") #

df_bhs1 <- df_bhs1 %>%
  mutate(across(starts_with("bhs1_"), ~ case_when(
    .x == 1 ~ "Agree",
    .x == 2 ~ "Disagree",
    TRUE ~ NA_character_
  ), .names = "{.col}_label"))

Base R alternative (no tidyverse required): #

# Select all columns starting with "bhs1_"
bhs_cols <- grep("^bhs1_", names(df_bhs1), value = TRUE)
# Batch recode and add new columns
df_bhs1[paste0(bhs_cols, "_recoded")] <- lapply(df_bhs1[bhs_cols], function(x) {
  ifelse(x == 1, 0, ifelse(x == 2, 1, NA))
})

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