Apply a Custom Function
Categorize employees into salary bands using a custom function. Available Data: DataFrame df with columns: name,...
pandasanalysistransformation
Intermediate
7m20 pts
0—
Python Practice Hub
pandas DataFrame Practice Questions
Select, group, merge and derive columns on a table in Python, without loops.
54 questionsFreeRuns in your browser
- loc / iloc
- Select rows and columns by label or position.
- groupby + agg
- Group rows and summarise each group.
- merge
- Join two DataFrames, like a SQL join.
What you will practice
- Select and filter with loc, iloc and boolean masks instead of chained indexing
- Aggregate with groupby().agg() and reshape the result with reset_index and pivot_table
- Join frames using merge with the right how= and on= arguments, and check the row count afterwards
- Replace row-by-row loops and apply() with vectorised column arithmetic and np.where
- Handle missing data with isna, fillna and dropna before the numbers reach a report
Read the full overview
A pandas DataFrame is a labelled, two-dimensional table in Python, and most analysis with it comes down to four moves: select the rows and columns you need, group and aggregate them, merge in another table, and derive new columns without looping. Doing that vectorised, with whole-column operations instead of a Python for loop, is what keeps the code both short and fast. These questions run real pandas in the browser against seeded DataFrames, so the output you check is the actual frame.
54 questionsPage 4 of 4
Standardize Matrix Columns (Broadcasting)
Given a 2D numpy array matrix (rows = observations, columns = features), compute the z-score-standardized version using...
numpyBroadcastingz-scoreVectorization+2Advanced12m30 pts0—Pairwise Euclidean Distance Matrix
Given a 2D numpy array points of shape (N, 2) representing N 2D points, compute the N×N pairwise Euclidean distance...
numpyBroadcastingVectorizationLinear Algebra+1Advanced15m35 pts0—Boolean Mask: Multi-Condition Filter
Given a 1D numpy array of float readings, extract a sorted array of all readings that are: Greater than 50, AND Less...
numpyBoolean MaskVectorizationFilteringIntermediate10m25 pts0—Cumulative Stats Along a Series
Given a 1D numpy array values of length N, compute three series and stack them into a 2D array of shape (N, 3): Column...
numpyVectorizationcumsumCumulative+1Advanced12m30 pts0—Funnel Step-Conversion Table
A mobile banking app's onboarding funnel needs a step-conversion breakdown for the weekly product review. You're given...
Product AnalyticsFunnelpandasConversion Rate+1Intermediate11m28 pts0—Cohort Retention Matrix in Pandas
You're given a small event log of user activity: each row is one (user, month) pair in which that user was active,...
Product AnalyticsCohort AnalysisRetentionpandas+1Advanced15m35 pts0—W3: Normalize an API Batch
Lowercase the plan codes in df and store the updated frame in result.
data-engineerw3pandasstandardizationBeginner5m10 pts0—W9: Partition Revenue Totals
Group df by load_date, summing mrr into partition_mrr with row counts n_rows. Store in result.
data-engineerw9pandasaggregationIntermediate10m10 pts0—
FAQ
pandas DataFrame practice: common questions
Short answers to what people get stuck on most. Tap a question to expand it.
How does pandas treat NaN in groupby sums and means?
Aggregations skip NaN by default, so a group still returns a total from its non-missing values.
- sum(), mean() and count() all ignore NaN.
- An all-NaN group sums to 0 unless you pass min_count=1 (shown here).
- Check df.isna().sum() first so you know how much data is being skipped.
SQL
df.groupby("city")["sales"].sum() # all-NaN group -> 0.0
df.groupby("city")["sales"].sum(min_count=1) # all-NaN group -> NaNWhat is the difference between loc and iloc?
loc selects by label, iloc selects by integer position.
- loc uses index values and column names, and slices include the end label.
- iloc uses positions, and slices exclude the end like normal Python.
- loc also takes boolean masks, the usual way to filter rows (shown here).
- Avoid chained indexing like df[mask]["col"] = x: it may assign to a copy (SettingWithCopyWarning).
SQL
df.loc[df["sales"] > 100, "city"] # label + boolean mask
df.iloc[0:2, 0] # first two rows, first columnHow do I join two DataFrames?
Use merge, which works like a SQL join. Choose how: inner, left, right or outer.
- Check the row count after merging: a many-to-many match multiplies rows.
- validate="many_to_one" makes pandas raise an error if the right key is not unique (shown here).
- Use pd.concat instead to stack DataFrames when there is no key to match on.
Python
orders.merge(
customers,
on="customer_id",
how="left",
validate="many_to_one",
)Related practice topics
Track your streak & earn XP
Sign up free to unlock your progress dashboard, daily streaks, and leaderboard ranking.
Full-Length Interview Practice Tests100% Free
10 role-based tests for Data Analysts, SQL Developers & Data Engineers with instant scorecards.