stack, unstack และ MultiIndex

Beyond melt and pivot, pandas can reshape using the index itself. The tools are stack and unstack on hierarchical labels. 🏗️

A MultiIndex stacks two or more label levels on one axis, so a single row can carry both a region and a year together.

stack moves the innermost column level into the row index, making the table taller and narrower in one move.

unstack is the reverse: it lifts an index level back into columns, making the table wider and shorter again.

Stack then unstack returns the original shape. Think of them as a paired toggle between row labels and column labels.

Pass a level name or number to choose which one shifts. This level control lets you reshape exactly the axis you mean.

Use loc with a tuple to drill into nested labels, grabbing one region-year combination directly from the stacked index.

The xs method slices one level cleanly, like every row for year 2021, without unpacking the whole index by hand.

To return to plain columns, call reset_index. It turns every index level back into an ordinary, exportable column.

By default stack quietly removes missing combinations. Set dropna to False when you need to keep every gap visible.

groupby with several keys returns a MultiIndex, so unstack is the natural finishing move to spread those groups into a grid.

Last check: stacking and unstacking.

You now wield pivot_table, melt, stack, and unstack. With layout under control, your analysis stays clean and flexible. 🎉