pandas value_counts() 方法对应的索引

Question

I often use value_counts() methods in pandas to get statistics.

For example, I can get the value_counts() result like below.

male     7825
female   6764

Is there any built-in function to get indices of the dataframe corresponding two labels(male and female).

Expected result: male_indices = [1,3,5,6,7, ..., 14589] , in which len(male_indices) = 7825

Answer 1

This is what groupby does. Consider the example dataframe df

np.random.seed([3,1415])
df = pd.DataFrame(dict(sex=np.random.choice(('male', 'female'), 10)))
print(df)

      sex
0    male
1  female
2    male
3  female
4    male
5    male
6  female
7    male
8  female
9  female

---

Use groupby.groups

df.groupby('sex').groups

{'female': Int64Index([1, 3, 6, 8, 9], dtype='int64'),
 'male': Int64Index([0, 2, 4, 5, 7], dtype='int64')}

Answer 2

Here's a minimal, somewhat-robust function that returns the indices corresponding to a given group within a given column in a DataFrame:

# create some data
d = pd.DataFrame({'sex': ['male', 'male', 'female', 'male', 'female', 'female', 'male'], 'age': [23, 24, 20, 32, 45, 43, 32]})

# returns a dictionary with group names as keys and indices corresponding 
# to those groups as values (can just use `list` or `set` to avoid pandas indexes
def get_indices(df, col):
    return {group: df[df[col] == group].index for group in set(df[col])}

# test it out
get_indices(d, 'sex')
Out[178]: 
{'female': Int64Index([2, 4, 5], dtype='int64'),
 'male': Int64Index([0, 1, 3, 6], dtype='int64')}