如何基于布尔值筛选存在True观测值的accountID列表

Got it, let's work through this. You need to pull all accountIDs that have at least one 'T' (True) in any of the date-specific var1 columns, and since you're dealing with millions of rows, we'll prioritize efficient, scalable methods over slower row-wise operations.

First, let's confirm your sample data (formatted for readability):

df1 <- data.frame(
  accountID = c(1, 2, 3, 4, 5),
  var1_2018_1 = c('T', 'F', 'F', 'T', 'T'),
  var1_2018_2 = c('F', 'F', 'F', 'T', 'T'),
  var1_2018_3 = c('T', 'F', 'F', 'T', 'T'),
  var1_2018_4 = c('T', 'F', 'F', 'T', 'T'),
  var1_2018_5 = c('T', 'F', 'F', 'T', 'T'),
  var1_2018_6 = c('T', 'F', 'F', 'T', 'T'),
  var1_2018_7 = c('T', 'F', 'F', 'T', 'T'),
  var1_2018_8 = c('F', 'F', 'T', 'F', 'T'),
  var1_2018_9 = c('T', 'F', 'F', 'T', 'T'),
  var1_2018_10 = c('T', 'F', 'F', 'T', 'T'),
  var1_2018_11 = c('T', 'F', 'F', 'T', 'T'),
  var1_2018_12 = c('T', 'F', 'F', 'T', 'T')
)

Method 1: Base R (Fastest for Large Datasets) #

This uses native R vector operations, which are optimized for speed—perfect for million-row datasets.

# Identify all columns starting with "var1_"
var_columns <- grep("^var1_", names(df1))

# Calculate the number of 'T's per row, filter rows with at least one, then extract accountID
valid_accounts <- df1$accountID[rowSums(df1[, var_columns] == "T") > 0]

# Output: [1] 1 3 4 5
print(valid_accounts)

Notes: #

• If your data has NA values, add na.rm = TRUE to rowSums to avoid invalid results:

  valid_accounts <- df1$accountID[rowSums(df1[, var_columns] == "T", na.rm = TRUE) > 0]
  

Method 2: dplyr (Tidyverse-Friendly) #

If you prefer the tidyverse syntax, use this vectorized approach (avoid rowwise() for large data—it's slower):

library(dplyr)

valid_accounts <- df1 %>%
  # Keep only accountID and var1 columns
  select(accountID, starts_with("var1_")) %>%
  # Filter rows where at least one var1 column is 'T'
  filter(rowSums(select(., -accountID) == "T") > 0) %>%
  # Pull the accountID column as a vector
  pull(accountID)

# Output: [1] 1 3 4 5
print(valid_accounts)

Why avoid rowwise()? #

Row-wise operations iterate over every row individually, which becomes a bottleneck with millions of records. The rowSums approach is fully vectorized, so it processes the entire dataset in bulk—way faster.

Expected Result #

For your sample data, the valid accountIDs are 1, 3, 4, 5 (account 2 has no 'T's in any var1 column).

内容的提问来源于stack exchange，提问作者TKTK