rolling mean values to fill NaN

Question

I got this function for fill main_df 's NaN cells. After had calculate the progressive rolling mean from every row, I'd then pick the value from the mean stream at each level of NaNs in order to substitute with those NaNs in the main_df cells.

Here is my trial:

main_df = pd.DataFrame({'a':[1,1,2,3,4,56,6],'b':[2,3,2,np.nan,4,3,6],'c':[1,1,2,3,4,56,np.nan]})

for i in range(len(main_df)):
    main_df.iloc[i] = main_df.iloc[i].fillna(value=main_df.iloc[i].ewm(span=1).mean())

Any suggestions to let it works?

Answer 1

I had final doublcheck my function works: I got substitution of nan in 'c' with values of ema2 at same level of nan

main_df = pd.DataFrame({'a':[1,1,2,3,4,56,6,4],'b':[2,3,2,np.nan,4,3,6,1],'c_copy':[1,1,2,3,4,56,np.nan,1],'c':[1,1,2,3,4,56,np.nan,1]})
main_df['ema2_c']= main_df['c'].ewm(span=2).mean()
main_df['c'] = main_df['c'].fillna(value=main_df['c'].ewm(span=2).mean())