Suppose I have the following dataframe:
#!/usr/bin/env python
import pandas as pd
df = pd.DataFrame([(1, 2, 1),
(1, 2, 2),
(1, 2, 3),
(4, 1, 612),
(4, 1, 612),
(4, 1, 1),
(3, 2, 1),
],
columns=['groupid', 'a', 'b'],
index=['India', 'France', 'England', 'Germany', 'UK', 'USA',
'Indonesia'])
print(df)
which gives:
groupid a b
India 1 2 1
France 1 2 2
England 1 2 3
Germany 4 1 612
UK 4 1 612
USA 4 1 1
Indonesia 3 2 1
This step might not be necessary / be different than how I imagine it. I'm actually only interested in Step 2, but having this helps me to think about it and explain what I want.
I want to group the data by groupid ( df.groupby(df['groupid'])
) and get something like this:
groupid a b
1 [2] [1, 2, 3]
4 [1] [612, 1]
3 [2] [1]
Then I want to find all group IDs which have only one entry in column b and for which the entry is equal to 1
.
Similarly, I want to find all group IDs which have either multiple entries or one entry which is not 1
.
You can compare set
s and then get values of index to list
s:
mask = df.groupby('groupid')['b'].apply(set) == set([1])
print (mask)
groupid
1 False
3 True
4 False
Name: b, dtype: bool
i = mask.index[mask].tolist()
print (i)
[3]
j = mask.index[~mask].tolist()
print (j)
[1, 4]
For new column use map
:
df['new'] = df['groupid'].map(df.groupby('groupid')['b'].apply(set) == set([1]))
print (df)
groupid a b new
India 1 2 1 False
France 1 2 2 False
England 1 2 3 False
Germany 4 1 612 False
UK 4 1 612 False
USA 4 1 1 False
Indonesia 3 2 1 True
old solution:
You can use transform
with nunique
for new Series
with same size as original df, so is possible compare it with 1
for uniqueness and then chain another condition for compare with 1
:
mask = (df.groupby('groupid')['b'].transform('nunique') == 1) & (df['b'] == 1)
print (mask)
India False
France False
England False
Germany False
UK False
USA False
Indonesia True
Name: b, dtype: bool
For unique values in list
s:
i = df.loc[mask, 'groupid'].unique().tolist()
print (i)
[3]
j = df.loc[~mask, 'groupid'].unique().tolist()
print (j)
[1, 4]
Detail:
print (df.groupby('groupid')['b'].transform('nunique'))
India 3
France 3
England 3
Germany 2
UK 2
USA 2
Indonesia 1
Name: b, dtype: int64
IIUC you can apply list and check for length using .str ie
temp = df.groupby('groupid')['b'].apply(list).to_frame()
temp
b
groupid
1 [1, 2, 3]
3 [1]
4 [612, 612, 1]
mask = (temp['b'].str.len() == 1) & (temp['b'].str[0] == 1)
temp[mask].index.tolist()
#[3]
temp[~mask].index.tolist()
#[1, 4]
I would go with
#group by the group id and than apply count for how many b entries are equal to 1
groups = df.groupby("groupid").apply(lambda group:len([x for x in
group["b"].values.tolist() if x == 1]))
#keep the groups containing 1 b equal to 1
groups = groups[groups == 1]
#print the indecies of the result (the groupid values)
print groups.index.values
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.