如何在R中计算给定双角色条件下其他角色的条件概率？

Great question! The bindata::condprob() function works perfectly for single-role conditional probabilities, but when you need to calculate probabilities given two simultaneous roles, we can build a custom solution that matches the clean output style you're used to. Here's a step-by-step breakdown:

Step 1: Reproduce Your Sample Data #

First, let's confirm we're working with the same dataset you provided:

# Recreate your sample data
SOdata <- structure(list(Role1 = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1), 
                         Role2 = c(0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1), 
                         Role3 = c(1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 1, 1), 
                         Role4 = c(0, 1, 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0), 
                         Role5 = c(0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1)), 
                    row.names = c(NA, -20L), class = c("tbl_df", "tbl", "data.frame"))

Step 2: Custom Function for Dual-Condition Probabilities #

We'll write a function that:

1. Filters your dataset to only include respondents who hold both target roles
2. Calculates the conditional probability of each remaining role (using the mean of binary variables, which equals the probability of being 1)
3. Formats the output to match condprob()'s structure

# Custom function for two-condition conditional probabilities
condprob_two <- function(data, cond_role1, cond_role2) {
  # Filter to respondents with both roles
  filtered_data <- data[data[[cond_role1]] == 1 & data[[cond_role2]] == 1, ]
  
  # Handle edge case: no respondents meet both conditions
  if (nrow(filtered_data) == 0) {
    stop(paste("No respondents hold both", cond_role1, "and", cond_role2))
  }
  
  # Calculate probabilities for other roles (mean = probability for binary vars)
  prob_calc <- colMeans(filtered_data[, !names(filtered_data) %in% c(cond_role1, cond_role2)])
  
  # Format output to match condprob() style
  result_df <- as.data.frame(t(prob_calc))
  rownames(result_df) <- paste0(cond_role1, "=1 & ", cond_role2, "=1")
  
  return(result_df)
}

Step 3: Test the Function #

Let's test it with your example: calculating probabilities for Role3, Role4, Role5 given that a respondent holds both Role1 and Role2:

# Calculate probabilities for Role1 + Role2 condition
condprob_two(SOdata, "Role1", "Role2")

Sample Output: #

Role3      Role4      Role5
Role1=1 & Role2=1 0.2352941 0.2352941 0.1176471

Step 4: Scale to All Role Pairs #

If you want to calculate probabilities for every possible pair of roles, you can automate the process with a loop:

# Get all role names
roles <- names(SOdata)

# Generate all unique role pairs
role_pairs <- t(combn(roles, 2))

# Calculate probabilities for every pair
all_pair_probs <- lapply(1:nrow(role_pairs), function(i) {
  cond1 <- role_pairs[i, 1]
  cond2 <- role_pairs[i, 2]
  condprob_two(SOdata, cond1, cond2)
})

# Combine results into a single data frame
all_pair_probs_df <- do.call(rbind, all_pair_probs)
print(all_pair_probs_df)

Key Note #

This approach works because for binary variables, the mean of the variable in the filtered subset directly equals the conditional probability of that role being held (since mean = number of 1s / total observations in the subset).

内容的提问来源于stack exchange，提问作者lawyeR