如何将 function 应用于具有重复索引的 dataframe

Question

Hello, I have the following dataframe with repetitive indices:

    Vol
1   25
1   15
2   20
2   30
3   25
3   10
4   15
4   20

I want to find a way that I can apply a function to do the sum of values in 'vol' column per each group of two indices. example:

1. take the first two indices using code:

one = df.loc[1:2]
one

it outputs first two indices 1,1 and 2,2:

    Vol
1   25
1   15
2   20
2   30

then do the sum of numbers in the column using:

one.sum()

output: 90

2. Repeat the process again but this time with index 2,2, and 3,3.

two=df.loc[2:3]
b=two.sum()
b

output: 85

3. Repeat the process again but this time with index 3,3 and 4,4

three=df.loc[3:4]
c=three.sum()
c

output : 70

I then put the output values in a new dataframe as

      vol2
0      90
1      85
2      70

As you can see, the method is very tedious especially with huge dataframes. Is there a way I can apply a function that runs through the dataframe as described above?

Answer 1

We need to sum the duplicate row first, the rolling sum

x.sum(level=0).rolling(2).sum().dropna()
Out[79]: 
    Vol
2  90.0
3  85.0
4  70.0

Answer 2

you can use this -

df.reset_index().set_axis(['index','Vol'], axis=1).groupby(['index'])['Vol'].sum().rolling(2).sum().dropna()

index
2    90.0
3    85.0
4    70.0

Answer 3

I don't know of a pure pandas way of doing this, but a little python can do the trick:

values = []
for x in range( df.index[0] , df.index[-1] ):
  value = df.loc[ x : x + 1 ].sum()
  values.append(value)

totals = pd.DataFrame( values )
print( totals )

If you are a fan of list comprehensions:

other = pd.DataFrame( [ df.loc[ x: x+ 1 ].sum() for x in range( df.index[0] , df.index[-1] ) ] )
print( other ) 

Both yield:

   vol
0   90 
1   85
2   70