R语言报错：non-numeric argument to binary operator 解决方案咨询

问题描述 #

运行代码时触发错误：Error in coefficients * c(x["price"], as.numeric(x["taste"] == "Sour" | : non-numeric argument to binary operator，尝试使用x$price但因x是原子向量无法生效。

原代码如下：

# Set the seed for reproducibility
set.seed(123)

# Create a design matrix with different levels for each factor
price <- rep(c(1, 2, 3), each = 4)
taste <- rep(c("Sweet", "Sour", "Salty", "Spicy"), times = 3)
texture <- rep(c("Crunchy", "Smooth"), each = 6)

design_matrix <- data.frame(price, taste, texture)

# Generate random utility coefficients for each factor
coefficients <- c(-0.5, -0.3, 0.8)

# Calculate the utility of each food option for each combination of factors
utility <- apply(design_matrix, 1, function(x) {
  sushi_utility <- sum(coefficients * c(x["price"], 
                                        as.numeric(x["taste"] == "Sour" | x["taste"] == "Salty"), 
                                        as.numeric(x["texture"] == "Smooth")))
  pizza_utility <- sum(coefficients * c(x["price"], 
                                        as.numeric(x["taste"] == "Sweet" | x["taste"] == "Spicy"), 
                                        as.numeric(x["texture"] == "Crunchy")))
  burger_utility <- sum(coefficients * c(x["price"], 
                                         as.numeric(x["taste"] == "Salty" | x["taste"] == "Spicy"), 
                                         as.numeric(x["texture"] == "Crunchy")))
  
  # Calculate the probability of choosing each food option
  probabilities <- exp(c(sushi_utility, pizza_utility, burger_utility))
  probabilities <- probabilities / sum(probabilities)
  
  # Simulate the choice of food option based on the probabilities
  food <- sample(c("Sushi", "Pizza", "Burger"), size = 1, prob = probabilities)
  
  # Return a data frame with the factors and the chosen food option
  data.frame(price = x["price"], taste = x["taste"],
             texture = x["texture"], food = food)
})

# Print the first few rows of the simulated dataset
head(utility)

错误原因 #

使用apply(design_matrix, 1, function(x))时，每一行的x会被转换为原子向量。由于原数据框包含字符型列（taste），整个向量的类型会被统一为字符型，导致x["price"]返回字符值，无法与数值型的coefficients进行乘法运算。

解决办法 #

方法1：在apply函数内转换向量类型 #

将原子向量x转换为数据框，保留各列的原始类型，避免类型统一：

utility <- apply(design_matrix, 1, function(x) {
  # 将原子向量转为数据框，保留原列类型
  x_df <- as.data.frame(t(x), stringsAsFactors = FALSE)
  
  sushi_utility <- sum(coefficients * c(x_df$price, 
                                        as.numeric(x_df$taste == "Sour" | x_df$taste == "Salty"), 
                                        as.numeric(x_df$texture == "Smooth")))
  pizza_utility <- sum(coefficients * c(x_df$price, 
                                        as.numeric(x_df$taste == "Sweet" | x_df$taste == "Spicy"), 
                                        as.numeric(x_df$texture == "Crunchy")))
  burger_utility <- sum(coefficients * c(x_df$price, 
                                         as.numeric(x_df$taste == "Salty" | x_df$taste == "Spicy"), 
                                         as.numeric(x_df$texture == "Crunchy")))
  
  probabilities <- exp(c(sushi_utility, pizza_utility, burger_utility))
  probabilities <- probabilities / sum(probabilities)
  
  food <- sample(c("Sushi", "Pizza", "Burger"), size = 1, prob = probabilities)
  
  data.frame(price = x_df$price, taste = x_df$taste,
             texture = x_df$texture, food = food)
})

或者更简洁地直接将x["price"]转为数值型（需每次提取时都转换）：

sushi_utility <- sum(coefficients * c(as.numeric(x["price"]), 
                                      as.numeric(x["taste"] == "Sour" | x["taste"] == "Salty"), 
                                      as.numeric(x["texture"] == "Smooth")))

方法2：使用dplyr::rowwise()替代apply（推荐） #

用dplyr的行处理方式，无需担心类型转换问题，代码更易读维护：

library(dplyr)

set.seed(123)

price <- rep(c(1, 2, 3), each = 4)
taste <- rep(c("Sweet", "Sour", "Salty", "Spicy"), times = 3)
texture <- rep(c("Crunchy", "Smooth"), each = 6)

design_matrix <- data.frame(price, taste, texture)

coefficients <- c(-0.5, -0.3, 0.8)

utility <- design_matrix %>%
  rowwise() %>%
  mutate(
    # 计算各食物效用
    sushi_utility = sum(coefficients * c(price, 
                                         as.numeric(taste %in% c("Sour", "Salty")), 
                                         as.numeric(texture == "Smooth"))),
    pizza_utility = sum(coefficients * c(price, 
                                         as.numeric(taste %in% c("Sweet", "Spicy")), 
                                         as.numeric(texture == "Crunchy"))),
    burger_utility = sum(coefficients * c(price, 
                                          as.numeric(taste %in% c("Salty", "Spicy")), 
                                          as.numeric(texture == "Crunchy"))),
    # 计算选择概率
    total_prob = exp(sushi_utility) + exp(pizza_utility) + exp(burger_utility),
    prob_sushi = exp(sushi_utility) / total_prob,
    prob_pizza = exp(pizza_utility) / total_prob,
    prob_burger = exp(burger_utility) / total_prob,
    # 模拟选择
    food = sample(c("Sushi", "Pizza", "Burger"), size = 1, prob = c(prob_sushi, prob_pizza, prob_burger))
  ) %>%
  select(price, taste, texture, food)

head(utility)

内容的提问来源于stack exchange，提问作者Ruggiero Rippo