Row-wise average for a subset of columns with missing values

I've got a 'DataFrame` which has occasional missing values, and looks something like this:

          Monday         Tuesday         Wednesday
================================================
Mike        42             NaN               12
Jenna       NaN            NaN               15
Jon         21              4                 1

I'd like to add a new column to my data frame where I'd calculate the average across all columns for every row.

Meaning, for Mike, I'd need (df['Monday'] + df['Wednesday'])/2, but for Jenna, I'd simply use df['Wednesday amt.']/1

Does anyone know the best way to account for this variation that results from missing values and calculate the average?

144661 次浏览

小开

最佳答案

You can simply:

df['avg'] = df.mean(axis=1)


Monday  Tuesday  Wednesday        avg
Mike       42      NaN         12  27.000000
Jenna     NaN      NaN         15  15.000000
Jon        21        4          1   8.666667

because .mean() ignores missing values by default: see docs.

To select a subset, you can:

df['avg'] = df[['Monday', 'Tuesday']].mean(axis=1)


Monday  Tuesday  Wednesday   avg
Mike       42      NaN         12  42.0
Jenna     NaN      NaN         15   NaN
Jon        21        4          1  12.5

小开

Alternative - using iloc (can also use loc here):

df['avg'] = df.iloc[:,0:2].mean(axis=1)

小开

Resurrecting this Question because all previous answers currently print a Warning.

In most cases, use assign():

df = df.assign(avg=df.mean(axis=1))

For specific columns, one can input them by name:

df = df.assign(avg=df.loc[:, ["Monday", "Tuesday", "Wednesday"]].mean(axis=1))

Or by index, using one more than the last desired index as it is not inclusive:

df = df.assign(avg=df.iloc[:,0:3]].mean(axis=1))