R语言变量创建两类问题：缺失值处理与多条件逻辑实现

Hey there! Let's work through your two R variable creation questions, making sure we handle those coded missing values (-99, -98) and actual NAs correctly—since leaving them unaddressed can mess up your binary variables.

1. Fixing Missing Values for the new Binary Variable #

Your original code df0$new <- ifelse(df0$old=="yes",1,0) will incorrectly assign 0 to any -99, -98, or NA values in old, treating them as "no" when they're actually missing data. Here's how to fix this properly:

First, it's best practice to convert those coded missing values to R's native NA—this makes all missing value handling consistent across functions:

# Recode -99, -98, and existing NA to R's NA
df0$old <- ifelse(df0$old %in% c(-99, -98, NA), NA, df0$old)

Then, you have a few clean options to create new:

• Base R shortcut: Convert a logical vector to integers (since TRUE becomes 1, FALSE becomes 0, and NA stays NA):

  df0$new <- as.integer(df0$old == "yes")
  

• Explicit ifelse: If you prefer more clarity:

  df0$new <- ifelse(df0$old == "yes", 1, ifelse(is.na(df0$old), NA, 0))
  

• dplyr case_when: Great for readability if you're using the tidyverse:

  library(dplyr)
  df0 <- df0 %>%
    mutate(new = case_when(
      old == "yes" ~ 1,
      !is.na(old) ~ 0,
      TRUE ~ NA_integer_ # Catch-all for missing values
    ))
  

2. Creating z with OR Logic + Missing Value Handling #

Your goal is to set z=1 if any of q1-q5 equals 1, z=0 only if all are non-missing and none equal 1, and z=NA if there are missing values but no 1s. Here's how to do this right:

First, recode those coded missing values in q1-q5 to NA just like we did for old:

# Recode -99/-98 to NA across all q1-q5 columns
df0[, paste0("q", 1:5)] <- lapply(df0[, paste0("q", 1:5)], function(x) {
  ifelse(x %in% c(-99, -98, NA), NA, x)
})

Now, two solid approaches for the OR logic:

• Base R with rowSums:

  # First set z=1 if any q equals 1 (ignoring NA for the check)
  df0$z <- as.integer(rowSums(df0[, paste0("q", 1:5)] == 1, na.rm = TRUE) > 0)
  # Fix cases where all q1-q5 are NA (set z to NA instead of 0)
  df0$z <- ifelse(rowSums(is.na(df0[, paste0("q", 1:5)])) == 5, NA, df0$z)
  

• dplyr if_any (cleanest!): The if_any function was built exactly for this "any column meets condition" scenario:

  df0 <- df0 %>%
    mutate(z = case_when(
      if_any(starts_with("q"), ~ .x == 1) ~ 1, # Any q equals 1
      if_all(starts_with("q"), ~ !is.na(.x)) ~ 0, # All q are non-missing and none are 1
      TRUE ~ NA_integer_ # Missing values present, no q equals 1
    ))
  

内容的提问来源于stack exchange，提问作者bvowe