R & dplyr：为组内选定成员分配组级别特征

Hey there! Let's work through how to assign group-level features to each individual in your dataset. From your description, you've got a large dataset split into contiguous groups (via grp), each with IDs starting at 1, plus is_child flags and momloc (0 or a group member's ID for the mother). Here are practical solutions using two common data analysis tools:

Using Python (Pandas) #

1. Basic Group-Level Summary Features #

If you want to assign aggregate stats like group size, proportion of children in the group, or average age to every individual in their group, this straightforward approach works:

import pandas as pd

# Example dataset matching your structure
data = pd.DataFrame({
    'grp': [1,1,1,1,2,2,2],
    'id': [1,2,3,4,1,2,3],
    'is_child': [False, True, True, False, True, False, True],
    'momloc': [0, 1, 1, 0, 2, 0, 2],
    'age': [30, 5, 3, 35, 4, 28, 2]
})

# Calculate group-level stats
group_stats = data.groupby('grp').agg(
    group_size=('id', 'count'),
    child_proportion=('is_child', 'mean'),
    avg_group_age=('age', 'mean')
).reset_index()

# Merge stats back to every individual in the group
data_with_group_features = pd.merge(data, group_stats, on='grp', how='left')

2. Mother-Derived Group-Level Features #

If you need to pull attributes from a child's mother (using momloc), we'll handle this within each group since IDs are unique per group:

def add_mother_attributes(group):
    # Create a lookup map from group ID to individual features
    id_lookup = group.set_index('id')[['age', 'is_child']].to_dict('index')
    # Fetch mother's features where momloc isn't 0
    group['mom_age'] = group['momloc'].apply(
        lambda x: id_lookup[x]['age'] if x != 0 else None
    )
    group['mom_is_child'] = group['momloc'].apply(
        lambda x: id_lookup[x]['is_child'] if x != 0 else None
    )
    return group

# Apply the function to each group
data_with_mom_features = data.groupby('grp', group_keys=False).apply(add_mother_attributes)

Using R (dplyr) #

1. Basic Group-Level Summary Features #

With dplyr, you can compute and assign group stats in one pipeline:

library(dplyr)

# Example dataset
data <- tibble(
  grp = c(1,1,1,1,2,2,2),
  id = c(1,2,3,4,1,2,3),
  is_child = c(FALSE, TRUE, TRUE, FALSE, TRUE, FALSE, TRUE),
  momloc = c(0, 1, 1, 0, 2, 0, 2),
  age = c(30,5,3,35,4,28,2)
)

# Add group-level features to each row
data_with_group_features <- data %>%
  group_by(grp) %>%
  mutate(
    group_size = n(),
    child_proportion = mean(is_child),
    avg_group_age = mean(age)
  ) %>%
  ungroup()

2. Mother-Derived Features #

To pull a mother's attributes using momloc, we can match IDs within each group:

data_with_mom_features <- data %>%
  group_by(grp) %>%
  mutate(
    mom_age = case_when(
      momloc != 0 ~ age[match(momloc, id)],
      TRUE ~ NA_real_
    ),
    mom_is_child = case_when(
      momloc != 0 ~ is_child[match(momloc, id)],
      TRUE ~ NA_logical_
    )
  ) %>%
  ungroup()

Quick Notes #

• Since your groups are already contiguous, the grouping functions will work as expected. If you ever need to confirm, sort the data by grp first (e.g., data.sort_values('grp') in pandas, data %>% arrange(grp) in R).
• Double-check that non-zero momloc values always correspond to valid IDs in the same group. You can add a quick validation step (e.g., in pandas: assert (data.groupby('grp')['momloc'].apply(lambda x: x[x!=0].isin(x.id).all())).all()).

内容的提问来源于stack exchange，提问作者andrewH