For example, I have two dataframes, df1 and df2.
I have to do the mean value between then. But, if df1 is np.nan, the mean will be the value from df2.
df1 = pd.DataFrame([1, 5, np.NaN, 4, np.NaN, 9])
df2 = pd.DataFrame([8, 2, np.NaN, np.NaN, 5, np.NaN])
I expect something like this.
df3 = pd.DataFrame([4.5, 3.5, np.NaN, 4, 5, 9])
My problem is, how can I do this condition?
You have several options. If you consider your single column DataFrames as Series, you could simply concat and mean on axis=1 (which ignores the NaNs by default):
out = pd.concat([df1[0], df2[0]], axis=1).mean(axis=1)
Output:
0 4.5
1 3.5
2 NaN
3 4.0
4 5.0
5 9.0
dtype: float64
If you potentially have multiple columns, you could manually compute the mean by computing the sum and dividing by the number of non-NaN:
out = df1.add(df2, fill_value=0).div(df1.notna().astype(int)+df2.notna())
Or, for a more generic option that could also handle an arbitrary number of DataFrames as input, using concat+groupy.mean:
out = pd.concat([df1, df2]).groupby(level=0).mean()
Output:
0
0 4.5
1 3.5
2 NaN
3 4.0
4 5.0
5 9.0
Using one column for each of the two DataFrames and a random choice of numbers or NaN.
