Wrangling & cleaning — quiz

mean() on 100 values with 40 missing divides by:

1. 100
2. 60
3. 40
4. It errors

Answer: 60 — Missing values are skipped entirely.

df.dropna() with no arguments removes:

1. Empty columns
2. Any row containing at least one NaN
3. Only fully empty rows
4. Duplicates

Answer: Any row containing at least one NaN — On a wide table this can remove most of the data.

To keep the most recent record per email:

1. drop_duplicates() alone
2. Sort by date, then drop_duplicates(keep="last")
3. dropna()
4. groupby(email)

Answer: Sort by date, then drop_duplicates(keep="last") — Without sorting, which row survives is arbitrary.

duplicated(keep=False) marks:

1. The first copy only
2. Every copy including the first
3. Nothing
4. Only nulls

Answer: Every copy including the first — Useful for inspecting all duplicates together.

Why does str.contains() fail on some columns?

1. Wrong dtype
2. The column contains NaN — pass na=False
3. Regex not enabled
4. Column too long

Answer: The column contains NaN — pass na=False — Missing values raise unless you tell it how to treat them.

str.split(" ", n=1, expand=True) returns:

1. A list
2. A DataFrame of the split parts
3. A string
4. A Series of lists

Answer: A DataFrame of the split parts — expand=True gives columns you can assign directly.

errors="coerce" does what to unparseable values?

1. Raises an error
2. Converts them to NaN
3. Leaves them as text
4. Drops the row

Answer: Converts them to NaN — Which lets the rest of the column convert.

Why pass format= to to_datetime?

1. It is faster only
2. To avoid ambiguous day/month misparsing
3. It is required
4. To handle nulls

Answer: To avoid ambiguous day/month misparsing — 03/04/2024 is ambiguous, and pandas may guess inconsistently.