How to use .loc with groupby and two conditions in pandas

Question

I asked a similar question here, but I want to expand on this question because I'm asked to do something a little different where I cannot use .duplicates()

I have a df that's grouped by 'Key'. I want to flag any row within a group where the discharge date matches the admit date AND between those rows, the row with the discharge date has a num1 value in the range of 5-12.

df =  pd.DataFrame({'Key': ['10003', '10003', '10003', '10003', '10003','10003','10034', '10034'], 
   'Num1': [12,13,13,13,12,13,15,12],
   'Num2': [121,122,122,124,125,126,127,128],
  'admit': [20120506, 20120508, 20121010,20121010,20121010,20121110,20120520,20120520],  'discharge': [20120508, 20120510, 20121012,20121016,20121023,20121111,20120520,20120520]})
df['admit'] = pd.to_datetime(df['admit'], format='%Y%m%d')
df['discharge'] = pd.to_datetime(df['discharge'], format='%Y%m%d')

initial df

    Key     Num1    Num2    admit       discharge
0   10003   12      121     2012-05-06  2012-05-08
1   10003   13      122     2012-05-08  2012-05-10
2   10003   13      122     2012-10-10  2012-10-12
3   10003   13      124     2012-10-10  2012-10-16
4   10003   12      125     2012-10-10  2012-10-23
5   10003   13      126     2012-11-10  2012-11-11
6   10034   15      127     2012-05-20  2012-05-20
7   10034   12      128     2012-05-20  2012-05-20

final df

    Key     Num1    Num2    admit       discharge   flag
0   10003   12      121     2012-05-06  2012-05-08  1
1   10003   13      122     2012-05-08  2012-05-10  1
2   10003   13      122     2012-10-10  2012-10-12  0
3   10003   13      124     2012-10-10  2012-10-16  0
4   10003   12      125     2012-10-10  2012-10-23  0
5   10003   13      126     2012-11-10  2012-11-11  0
6   10034   15      127     2012-05-20  2012-05-20  1
7   10034   12      128     2012-05-20  2012-05-20  1

I was trying to use filter() but I can't quite figure out how to apply any() to the discharge date. My logic was to pick the first admit date in a group and then check that date among each discharge date and once there is a match then check if the row that has the same discharge date has a value in Num1 in the range of 5-12.

num1_range = [5,6,7,8,9,10,11,12]
df.loc[df.groupby(['Key']).filter(lambda x : (x['admit'] == x['discharge'].any())&(x['Num1'].isin(num1_range).any())),'flag']=1

I'm getting an error

ValueError: cannot set a Timestamp with a non-timestamp

Answer 1

I believe you are looking for either of 2 conditions to be satisfied for flag = True:

1. Admit date is equal to any discharge date within the group (Key).
2. Discharge date is equal to any admit date within the group, provided Num1 is in the range 5 to 12 inclusive.

The below logic produces the result in line with your desired output.

Solution

d1 = df.groupby('Key')['admit'].apply(set).to_dict()
d2 = df.groupby('Key')['discharge'].apply(set).to_dict()

def flagger(row):
    match1, match2 = row['discharge'] in d1[row['Key']], row['admit'] in d2[row['Key']]
    return match2 or (match1 and (row['Num1'] in range(5, 13)))

df['flag'] = df.apply(flagger, axis=1).astype(int)

Result

     Key  Num1  Num2      admit  discharge  flag
0  10003    12   121 2012-05-06 2012-05-08     1
1  10003    13   122 2012-05-08 2012-05-10     1
2  10003    13   122 2012-10-10 2012-10-12     0
3  10003    13   124 2012-10-10 2012-10-16     0
4  10003    12   125 2012-10-10 2012-10-23     0
5  10003    13   126 2012-11-10 2012-11-11     0
6  10034    15   127 2012-05-20 2012-05-20     1
7  10034    12   128 2012-05-20 2012-05-20     1

Explanation

• Create 2 dictionary mapping Key -> Admit dates and Key -> Discharge dates respectively.
• Use these 2 dictionaries to apply the criteria specified by row using pd.DataFrame.apply.