如何按规则筛选DataFrame的Top 10样本？含分组限制与NA处理

Let's break down how to achieve your desired result step by step. First, let's recap the rules to make sure we're aligned:

• Keep all samples with group = NA that fall into the top value range
• For samples with non-NA groups, only keep 1 sample per group (the one with the highest value)
• First retain the specific samples: 36, 22, 2, 20, 11, then fill the remaining 5 spots with the next highest-value samples that follow the rules.

Step 1: Reproduce the Original Data #

First, let's generate the dataset as you provided:

set.seed(1)
df <- data.frame(
  sample = 1:50,
  value = runif(50),
  group = c(rep(NA, 20), gl(3, 10))
)

Step 2: Define Priority Samples & Sort the Data #

First, list the samples you want to prioritize, then sort the entire dataset by value in descending order (so we're always picking from the highest values first):

library(dplyr)

# Define the samples we need to keep first
priority_samples <- c(36, 22, 2, 20, 11)

# Sort the full dataset by value descending
df_sorted <- df %>% arrange(desc(value))

Step 3: Extract Priority Samples #

Pull out the priority samples first—these are guaranteed to be in our final result:

df_priority <- df_sorted %>% filter(sample %in% priority_samples)

Step 4: Process Remaining Samples (Follow Rules) #

For the remaining samples, we need to apply the group constraint:

• Keep all NA-group samples (since they don't have group limits)
• For non-NA groups, only keep the highest-value sample per group (since we already sorted, slice(1) gives us the top value for each group)

df_remaining <- df_sorted %>%
  filter(!sample %in% priority_samples) %>%  # Exclude already kept samples
  group_by(group) %>%
  slice(1) %>%  # Keep top 1 per group (NA groups are kept as-is)
  ungroup()

Step 5: Fill the Remaining Spots #

We need 5 more samples to reach 10 total. Grab the top 5 from our processed remaining samples:

df_additional <- df_remaining %>% slice_head(n = 5)

Step 6: Combine & Finalize #

Merge the priority samples and additional samples, then re-sort by value to get the final ordered result:

df_final <- bind_rows(df_priority, df_additional) %>%
  arrange(desc(value)) %>%
  slice_head(n = 10)  # Ensure we have exactly 10 samples

# View the final result
print(df_final)

Alternative One-Liner Approach #

If you prefer a more condensed version, you can mark priority samples and combine the logic:

df_final <- df %>%
  arrange(desc(value)) %>%
  mutate(is_priority = sample %in% priority_samples) %>%
  # First take all priority samples
  filter(is_priority) %>%
  bind_rows(
    # Then take top 1 per group from non-priority samples
    df %>%
      arrange(desc(value)) %>%
      filter(!sample %in% priority_samples) %>%
      group_by(group) %>%
      slice(1) %>%
      ungroup()
  ) %>%
  arrange(desc(value)) %>%
  slice_head(n = 10)

This will give you exactly the 10 samples you want: the 5 priority ones, plus the next 5 highest-value samples that adhere to the group uniqueness rule.

内容的提问来源于stack exchange，提问作者user42485