Merging dataframes (with different date time --monthly vs daily) at the same time applying lagged values to one of the dataframes

user_14459677 0
Viewed 39

I have 2 dataframes I wish to merge:

df1 looks like this:

Date          Col1  Col 2  Col 3    Col 4   
    
2016-03     27.57   0.93    28.7    1.57    
2016-04     25.83   0.23    28.34   0.84    
2016-05     24.55   0.27    27.11   0.03

df2 looks like this:

Date          ColA            

2016-03-21  7.640769230769231   
2016-03-22  7.739720279720279   
2016-03-23  7.577311827956988   
2016-03-24  7.745416666666666

As you can see, df1 is a monthly data and df2 is a daily data. However, I want to merge them in a daily format (following df2) but I also want df1 to be lagged (lag = -30)

This is my desired output:

Output:

   Date          ColA              Col1  Col 2  Col 3    Col 4     
      
    2016-03-21  7.640769230769231   25.83   0.23    28.34   0.84
    2016-03-22  7.739720279720279   25.83   0.23    28.34   0.84
    2016-03-23  7.577311827956988   25.83   0.23    28.34   0.84
    2016-03-24  7.745416666666666   25.83   0.23    28.34   0.84


....2016-04-01  xxxxxxxx            24.55   0.27    27.11   0.03

I tried this but, they just merge and the lags were not applied.

out = (df2.merge(df1.shift(-30), on='Date').axis=1)

EDIT: Since I can't make the below suggestions work on my specific problem, what I did is this:

   out['Col1']= out['Col1'].shift(7).dropna()
    out = pd.merge_asof(df1, df2,  on = 'Date')

This is to lag only 1 column (which I decided to do since most columns have possibility to have different lag times.

pandas lag concatenation merge python
2 Answers

For merge use merge_asof:

print (df1)
      Date   Col1  Col 2  Col 3  Col 4
0  2016-03  27.57   0.93  28.70   1.57
1  2016-04  25.83   0.23  28.34   0.84
2  2016-05  24.55   0.27  27.11   0.03

df1['Date'] = pd.to_datetime(df1['Date'])
print (df1)
        Date   Col1  Col 2  Col 3  Col 4
0 2016-03-01  27.57   0.93  28.70   1.57
1 2016-04-01  25.83   0.23  28.34   0.84
2 2016-05-01  24.55   0.27  27.11   0.03
    

np.random.seed(20)
df2 = pd.DataFrame({'Date': pd.date_range('2016-02-20', periods=20),
                    'ColA': np.random.randint(5, size=20)})
    
print (df2)
         Date  ColA
0  2016-02-20     3
1  2016-02-21     2
2  2016-02-22     4
3  2016-02-23     2
4  2016-02-24     1
5  2016-02-25     4
6  2016-02-26     3
7  2016-02-27     2
8  2016-02-28     0
9  2016-02-29     0
10 2016-03-01     2
11 2016-03-02     2
12 2016-03-03     3
13 2016-03-04     3
14 2016-03-05     0
15 2016-03-06     0
16 2016-03-07     1
17 2016-03-08     2
18 2016-03-09     2
19 2016-03-10     3

#default direction='backward'
df = pd.merge_asof(df2, df1, on='Date', tolerance=pd.Timedelta(7, 'days'))
print (df)
         Date  ColA   Col1  Col 2  Col 3  Col 4
0  2016-02-20     3    NaN    NaN    NaN    NaN
1  2016-02-21     2    NaN    NaN    NaN    NaN
2  2016-02-22     4    NaN    NaN    NaN    NaN
3  2016-02-23     2    NaN    NaN    NaN    NaN
4  2016-02-24     1    NaN    NaN    NaN    NaN
5  2016-02-25     4    NaN    NaN    NaN    NaN
6  2016-02-26     3    NaN    NaN    NaN    NaN
7  2016-02-27     2    NaN    NaN    NaN    NaN
8  2016-02-28     0    NaN    NaN    NaN    NaN
9  2016-02-29     0    NaN    NaN    NaN    NaN
10 2016-03-01     2  27.57   0.93   28.7   1.57
11 2016-03-02     2  27.57   0.93   28.7   1.57
12 2016-03-03     3  27.57   0.93   28.7   1.57
13 2016-03-04     3  27.57   0.93   28.7   1.57
14 2016-03-05     0  27.57   0.93   28.7   1.57
15 2016-03-06     0  27.57   0.93   28.7   1.57
16 2016-03-07     1  27.57   0.93   28.7   1.57
17 2016-03-08     2  27.57   0.93   28.7   1.57
18 2016-03-09     2    NaN    NaN    NaN    NaN
19 2016-03-10     3    NaN    NaN    NaN    NaN

df = pd.merge_asof(df2, df1, on='Date', 
                   tolerance=pd.Timedelta(7, 'days'), direction='forward')
print (df)
         Date  ColA   Col1  Col 2  Col 3  Col 4
0  2016-02-20     3    NaN    NaN    NaN    NaN
1  2016-02-21     2    NaN    NaN    NaN    NaN
2  2016-02-22     4    NaN    NaN    NaN    NaN
3  2016-02-23     2  27.57   0.93   28.7   1.57
4  2016-02-24     1  27.57   0.93   28.7   1.57
5  2016-02-25     4  27.57   0.93   28.7   1.57
6  2016-02-26     3  27.57   0.93   28.7   1.57
7  2016-02-27     2  27.57   0.93   28.7   1.57
8  2016-02-28     0  27.57   0.93   28.7   1.57
9  2016-02-29     0  27.57   0.93   28.7   1.57
10 2016-03-01     2  27.57   0.93   28.7   1.57
11 2016-03-02     2    NaN    NaN    NaN    NaN
12 2016-03-03     3    NaN    NaN    NaN    NaN
13 2016-03-04     3    NaN    NaN    NaN    NaN
14 2016-03-05     0    NaN    NaN    NaN    NaN
15 2016-03-06     0    NaN    NaN    NaN    NaN
16 2016-03-07     1    NaN    NaN    NaN    NaN
17 2016-03-08     2    NaN    NaN    NaN    NaN
18 2016-03-09     2    NaN    NaN    NaN    NaN
19 2016-03-10     3    NaN    NaN    NaN    NaN

df = pd.merge_asof(df2, df1, on='Date', 
                   tolerance=pd.Timedelta(7, 'days'), direction='nearest')
print (df)
         Date  ColA   Col1  Col 2  Col 3  Col 4
0  2016-02-20     3    NaN    NaN    NaN    NaN
1  2016-02-21     2    NaN    NaN    NaN    NaN
2  2016-02-22     4    NaN    NaN    NaN    NaN
3  2016-02-23     2  27.57   0.93   28.7   1.57
4  2016-02-24     1  27.57   0.93   28.7   1.57
5  2016-02-25     4  27.57   0.93   28.7   1.57
6  2016-02-26     3  27.57   0.93   28.7   1.57
7  2016-02-27     2  27.57   0.93   28.7   1.57
8  2016-02-28     0  27.57   0.93   28.7   1.57
9  2016-02-29     0  27.57   0.93   28.7   1.57
10 2016-03-01     2  27.57   0.93   28.7   1.57
11 2016-03-02     2  27.57   0.93   28.7   1.57
12 2016-03-03     3  27.57   0.93   28.7   1.57
13 2016-03-04     3  27.57   0.93   28.7   1.57
14 2016-03-05     0  27.57   0.93   28.7   1.57
15 2016-03-06     0  27.57   0.93   28.7   1.57
16 2016-03-07     1  27.57   0.93   28.7   1.57
17 2016-03-08     2  27.57   0.93   28.7   1.57
18 2016-03-09     2    NaN    NaN    NaN    NaN
19 2016-03-10     3    NaN    NaN    NaN    NaN
user_29010020

IIUC, you can use:

df1['Date'] = pd.to_datetime(df1['Date'])
df2['Date'] = pd.to_datetime(df2['Date'])

out = (df2.merge(df1.assign(Date=df1['Date'].sub(pd.DateOffset(months=1))
                                            .dt.to_period('M')),
                 left_on=df2['Date'].dt.to_period('M'), right_on='Date',
                 how='left', suffixes=(None, '_df1'))
      )

or:

out = (df2.merge(df1.assign(Date=df1['Date'].dt.to_period('M').sub(1)),
                 left_on=df2['Date'].dt.to_period('M'), right_on='Date',
                 how='left', suffixes=(None, '_df1'))
      )

output:

        Date      ColA Date_df1   Col1  Col 2  Col 3  Col 4
0 2016-03-21  7.640769  2016-03  25.83   0.23  28.34   0.84
1 2016-03-22  7.739720  2016-03  25.83   0.23  28.34   0.84
2 2016-03-23  7.577312  2016-03  25.83   0.23  28.34   0.84
3 2016-03-24  7.745417  2016-03  25.83   0.23  28.34   0.84
user_163434640
Related