pandas Series.cumsum（）vs pandas.expanding_sum（）

Question

assuming I have a pandas Series s, what is the difference between s.cumsum() and pd.expanding_sum(s)? (I guess the answer should be the same also for cummax()/cummin(), and pd.expanding_max()/pd.expanding_min())

The docs say:

Note The output of the rolling_ and expanding_ functions do not return a NaN if there are at least min_periods non-null values in the current window. This differs from cumsum, cumprod, cummax, and cummin, which return NaN in the output wherever a NaN is encountered in the input.

Is this the only difference?

(assuming this is the only difference I don't understand why there needs to be 2 different methods defined for this very similar functionality)

Answer 1

They are basically the same, but you will get NaNs with expanding_sum until you reach the required minimum number of observations.

s = pd.Series([1] * 5)

>>> s.cumsum()
0    1
1    2
2    3
3    4
4    5
dtype: int64

>>> pd.expanding_sum(s, min_periods=3)
0   NaN
1   NaN
2     3
3     4
4     5
dtype: float64

expanding_sum also allows you to pre-conform your time indexed data, apparently based on mean .

s = pd.Series([0, 1] * 5, index=pd.date_range('2015-1-1', periods=10, freq='12H'))

>>> s
2015-01-01 00:00:00    0
2015-01-01 12:00:00    1
2015-01-02 00:00:00    0
2015-01-02 12:00:00    1
2015-01-03 00:00:00    0
2015-01-03 12:00:00    1
2015-01-04 00:00:00    0
2015-01-04 12:00:00    1
2015-01-05 00:00:00    0
2015-01-05 12:00:00    1
Freq: 12H, dtype: int64

>>> pd.expanding_sum(s, min_periods=3, freq='1D')
2015-01-01    NaN
2015-01-02    NaN
2015-01-03    1.5
2015-01-04    2.0
2015-01-05    2.5
Freq: D, dtype: float64

The documentation contains more information on expanding window moment functions .

Regarding the difference in how the two methods treat NaNs, here is an illustrative example:

s = pd.Series([1] * 5)
s.loc[2] = None

>>> s.cumsum()
0     1
1     2
2   NaN
3     3
4     4
dtype: float64

>>> pd.expanding_sum(s)
0    1
1    2
2    2
3    3
4    4
dtype: float64