Effective Data Manipulation with pandas/

...

/

Stacking and Flattening Data

Stacking and Flattening Data

Explore stacking, unstacking, and flattening data.

We'll cover the following...

Unstacking

We’ve used the unstack method previously but not discussed it. It’s a powerful method (along with its complement, stack) to reshape our data. At a high level, unstack moves an index level into the columns. Usually, we use this operation on multi-index data, moving one of the indices into the columns (creating hierarchical columns).

Jupyter showing a transposed view of DataFrame. Notice that we see ten complete samples of data showing on the screen without scrolling.
Jupyter showing a transposed view of DataFrame. Notice that we see ten complete samples of data showing on the screen without scrolling.

The stack method does the reverse, moving a multilevel column into the index.

Python 3.8
import pandas as pd
url = 'https://github.com/mattharrison/datasets/raw/master/data/'\
      '2020-jetbrains-python-survey.csv'
jb = pd.read_csv(url)
jb2 = tweak_jb(jb)
print(jb2)

We’ll create a hierarchical index or multi-index by grouping with multiple columns. Let’s take the size of responses to are_you_datascientist column by country:

Python 3.8
print(jb2
  .groupby(['country_live', 'are_you_datascientist'])
  .size()
)

Notice that the result is a series with a multi-index. This result is useful but a little hard to scan through. It would be easier if we had countries in the index and each of the responses to are_you_datascientist as their own ...