How to do sum previous and next row with skipping Nan in Pandas-CodePudding

I have one column in my Dataframe and I am trying to calculate the energy loss with formula. Problem is that I want to use only two valid rows each time where values are not NaN. Energy is the input column and looking for something like loss column.

Energy	loss
NaN	Nan
NaN	Nan
NaN	Nan
4	Nan
NaN	Nan
3	1/2(4^2-3^2)
NaN	Nan
11	Nan
3	1/2(3^2-11^2)
NaN	NaN
14	Nan

I tried Lambda custom function but not able to send the next row.

CodePudding user response：

Try something like this:

df = pd.DataFrame({'Energy':[4,None,3,None,11,3,None,14]})
energy = df.Energy.dropna()
def my_loss(series):
    return 1/2*(series.iloc[0]**2-series.iloc[1]**2)
loss = energy.rolling(2).apply(my_loss)
df['loss'] = loss[1::2]  # skip half of the results

Basically you apply your custom function in the rolling of the droped nan energy, and then merge it again with your df.