基于双Likert量表因子构建新分类因子的R语言实现问题

Got it, let's fix this for you! You're trying to derive a new factor Z based on two 5-point Likert variables (X and Y), with specific rules for each category. The issue with your initial code is that you weren't actually mapping the values of X and Y to the Z categories—you just created a static factor with all three levels. Here's how to do it properly:

Step 1: Clarify the Logic #

First, let's restate your rules to make sure we're aligned:

• None: Both X and Y are "[1] Never"
• Mild: Y is "[1] Never", but X is anything else (Rarely/Sometimes/Often/Very Often)
• Severe: Y is anything other than "[1] Never" (regardless of X's value)

Step 2: Implement in R #

We'll cover two common approaches—base R and dplyr (for data frame workflows) — depending on what you're using.

Option 1: Base R (for standalone vectors or data frames) #

Assume X and Y are numeric vectors (coded 1-5). We'll use nested ifelse() to assign each case to the correct Z category, then convert to an ordered factor (important if you plan to use Z in statistical tests or visualizations):

# Create Z as a character vector first by applying your rules
Z <- ifelse(X == 1 & Y == 1, "None",
            ifelse(Y == 1, "Mild", "Severe"))

# Convert to a factor with explicit level order (matches your priority)
Z <- factor(Z, levels = c("None", "Mild", "Severe"), ordered = TRUE)

Option 2: dplyr (tidyverse workflow for data frames) #

If your data is in a data frame (e.g., df), case_when() makes the logic more readable and scalable:

library(dplyr)

df <- df %>%
  mutate(
    # Assign Z categories using clear conditionals
    Z = case_when(
      X == 1 & Y == 1 ~ "None",
      Y == 1 ~ "Mild",  # Catches all Y=1 cases that aren't X=1
      TRUE ~ "Severe"   # All remaining cases (Y != 1)
    ),
    # Convert Z to an ordered factor
    Z = factor(Z, levels = c("None", "Mild", "Severe"), ordered = TRUE)
  )

If X/Y are Factor Variables (not numeric) #

If X and Y are already factors with levels like c("Never", "Rarely", "Sometimes", "Often", "Very Often"), adjust the conditionals to check the level names instead:

# Base R version for factor X/Y
Z <- ifelse(as.character(X) == "Never" & as.character(Y) == "Never", "None",
            ifelse(as.character(Y) == "Never", "Mild", "Severe"))

Z <- factor(Z, levels = c("None", "Mild", "Severe"), ordered = TRUE)

Why This Works #

The key difference from your initial code is that we're dynamically assigning each observation to a Z category based on the actual values of X and Y, rather than just creating a static factor with all three levels. The ordered = TRUE argument ensures the factor levels follow the logical severity order (None → Mild → Severe), which is crucial for any downstream analysis.

内容的提问来源于stack exchange，提问作者Andreas