Welcome to Software Development on Codidact!
Will you help us build our independent community of developers helping developers? We're small and trying to grow. We welcome questions about all aspects of software development, from design to code to QA and more. Got questions? Got answers? Got code you'd like someone to review? Please join us.
How to compress columns of dataframe by function
Problem
How can I compress each column of a dataframe to the output of a function (i.e., mean), preserving columns?
MWE
import pandas as pd
data = {"A": [1, 2, 3, 4], "B": [5, 6, 7, 8]}
df = pd.DataFrame(data)
A B
0 1 5
1 2 6
2 3 7
3 4 8
Desired Output
A B
0 2.5 6.5
Tried
I was thinking one of the apply()
or aggregate()
functions would
work.
apply
has a results_type
field, but none of them produced the desired
output.
Workarounds
These are workarounds I figured out that produce the desired outcome, but I find them cumbersome and un-intuitive, and feel there must be a simpler way I have not discovered.
Repetitive, cumbersome, and not scalable:
df = pd.DataFrame({"A": [df["A"].mean()], "B": [df["B"].mean()]})
Un-intuitive and long:
df.mean().to_frame().transpose()
2 answers
Since pandas
uses numpy
for these computations under the hood, I would have suggested to use df.mean(keepdims=True)
, but apparently this has been explicitly disabled by pandas
.
However, after looking into the docs, I noticed you should be able to get the desired result as follows (note the []
):
>>> df.agg(["mean"])
A B
mean 2.5 6.5
The list can also contain functions and/or more operations. Note that this will introduce a row (with index) for each function.
0 comment threads
Conceptually, applying a function along an axis of a DataFrame
(i.e., applying it to each row or column) inherently produces a Series
: a two-dimensional result is collapsed to a one-dimensional result, because one-dimensional "lines" of data are fed into a function that produces a scalar value.
Such a series can be appended as a row to an existing DataFrame
, if the labels are compatible - such as with the original DataFrame
:
>>> df.loc['avg'] = df.mean()
>>> df
A B
0 1.0 5.0
1 2.0 6.0
2 3.0 7.0
3 4.0 8.0
avg 2.5 6.5
However, creating a row by itself - i.e., in a new DataFrame - either requires an existing DataFrame with those labels:
>>> x = df.mean()
>>> y = pd.DataFrame(columns=x.index)
>>> y.loc[0] = x
>>> y
A B
0 2.5 6.5
or creating it as you have tried already. As a hint, the T
property saves some typing:
>>> df.mean().to_frame().T
A B
0 2.5 6.5
However, there is not an option for converting a Series
directly to a single-row DataFrame
; it converts to a column regardless.
Reference: https://stackoverflow.com/questions/59406045
2 comment threads