CODEWITHSUNDEEP
pythonpandas
Anthony W
Anthony W

Reputation: 1327

Trying to generate a rolling_sum with Pandas

I have a simple dataFrame with regular samples:

dates = pd.date_range('20150101', periods=6)
df = pd.DataFrame([1,2,3,4,5,6], index=dates, columns=list('A'))
df.loc[:,'B'] = 0
df.iloc[0,1] =10
df


Out[119]:
            A   B
2015-01-01  1   10
2015-01-02  2   0
2015-01-03  3   0
2015-01-04  4   0
2015-01-05  5   0
2015-01-06  6   0

What I would like to do is using the rolling_sum function in Pandas to create values in column B such that I get the following table:

            A   B
2015-01-01  1   11
2015-01-02  2   13
2015-01-03  3   16
2015-01-04  4   20
2015-01-05  5   25
2015-01-06  6   31

That is, B(i+1) = A(i+1) + B(i) - note there is a special case at B(0) as there is an initial value of 10. I've tried using the following, but it gives the wrong answer: 

df.B = pd.rolling_sum(df.A + df.B, 1)

Upvotes: 2

Views: 57

Answers (1)

Warren Weckesser
Warren Weckesser

Reputation: 114976

Try expanding_sum:

In [15]: df
Out[15]: 
            A   B
2015-01-01  1  10
2015-01-02  2   0
2015-01-03  3   0
2015-01-04  4   0
2015-01-05  5   0
2015-01-06  6   0

In [16]: df.B = pd.expanding_sum(df.A) + df.B[0]

In [17]: df
Out[17]: 
            A   B
2015-01-01  1  11
2015-01-02  2  13
2015-01-03  3  16
2015-01-04  4  20
2015-01-05  5  25
2015-01-06  6  31

Upvotes: 2

Related Questions