Level 4 · Series or DataFrame?

Why this distinction matters

pandas has two main objects, and almost every confusing error in your first weeks comes from having one when you expected the other.

A Series is one column: values plus an index. It has one dtype and a name (the column name). Series have methods for working on values, such as .value_counts(), .unique(), .str.lower(), .dt.year, .map() and .between().

A DataFrame is a table of one or more columns that share an index. DataFrames have methods for working on tables, such as .merge(), .groupby() with several columns, .sort_values(by=...), .drop_duplicates(subset=...) and .to_csv().

The rule

orders["status"]             # Series
orders[["status"]]           # DataFrame, 1 column
orders[["status", "channel"]]   # DataFrame, 2 columns

The inner brackets create a Python list. Whenever you select with a list, you get a DataFrame, even if the list holds a single name.

Checking what you have

When in doubt, ask Python:

type(orders["status"])       # pandas.core.series.Series
type(orders[["status"]])     # pandas.core.frame.DataFrame
orders[["status"]].shape     # (12000, 1)
orders["status"].shape       # (12000,)   one dimension only

The shape is a good tell: a Series has a one-element shape, a DataFrame always has two.

A useful mental model

Think of a DataFrame as a dictionary of Series that all share the same index. Asking for one key (orders["status"]) hands you that one Series. Asking with a list of keys hands you a smaller dictionary, which is still a DataFrame. This also explains why adding a column looks like adding a dictionary key: orders["is_web"] = orders["channel"] == "web".

Converting between them

orders[["status"]].squeeze()        # one-column DataFrame to Series
orders["status"].to_frame()         # Series to one-column DataFrame

Where it bites

• orders[["status"]].value_counts() works but returns counts indexed by a one-level MultiIndex, which prints and merges differently from orders["status"].value_counts().
• .str and .dt accessors exist only on Series.
• Assigning a new column needs a Series (or a list or a single value) on the right-hand side, not a DataFrame.
• Many aggregation results are Series. orders.groupby("channel")["shipping_fee"].mean() is a Series indexed by channel; add .reset_index() to turn it into a two-column table.

Getting this right also makes task statements in this game easy to follow: when a level asks for a Series, return one column’s worth of values; when it asks for a DataFrame with named columns, return a table with exactly those columns.