Prepare for pandas interview questions by practicing how to choose an operation, explain the shape and labels it returns, and state the data assumptions behind it. The 51 questions below move from core objects and selection through cleaning, grouping, joins, reshaping, time series, and file handling.
These are practice prompts, not a frequency ranking. Use the examples to explain your reasoning, then adapt syntax to the pandas version installed for the role.
As an Amazon Associate I earn from qualifying purchases.
Fundamentals and inspection
1. What is pandas, and what data-analysis work does it support?
pandas is a Python library for working with labeled, tabular and time-oriented data. It provides structures and operations for selecting, cleaning, summarizing, combining, reshaping, and reading or writing datasets. Python is the programming language; pandas is one library used within it. The pandas User Guide organizes these workflows.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitches2. What is a Series?
A Series is a one-dimensional labeled array. It contains values and an associated index, so its elements can be selected and aligned by labels as well as handled in sequence.
#1 Best Overall
3. What is a DataFrame?
A DataFrame is a two-dimensional, size-mutable tabular structure with labeled rows and columns. Its columns may hold different data types, such as numbers, strings, or dates. See the DataFrame API reference.
4. How are a Series and a DataFrame related?
A DataFrame is made up of labeled columns, and selecting one column with a single label commonly returns a Series: df["revenue"]. Selecting multiple columns with a list, as in df[["revenue", "region"]], returns a DataFrame. Their main difference is dimensionality: one dimension for a Series, two for a DataFrame.
5. What is an index, and why do labels matter?
An index labels rows; column labels identify fields. Labels make it possible to select particular records and align objects by matching labels during many operations. Do not assume a label is the same as a row’s current position: an index can be nonconsecutive or contain strings.
6. How do you inspect a DataFrame before transforming it?
Check its dimensions, column names, data types, and representative rows before choosing an operation. For example, df.shape reports row and column counts, df.columns shows field labels, df.dtypes reports column types, and df.head() displays sample rows. These checks help reveal unexpected types, labels, or structure.
7. How do you inspect or change column types?
Inspect types with df.dtypes and decide what each field represents before converting it. A numeric-looking column might contain malformed text; a date stored as text may need parsing. Use an appropriate conversion, such as pd.to_datetime for dates or astype for a suitable dtype, then check the result and account for invalid values. No single conversion is right for every column.
Selection and indexing
8. How does label-based selection differ from positional selection?
.loc selects by labels, while .iloc selects by integer position. For instance, df.loc["row_a", "sales"] asks for the value at a named row and column; df.iloc[0, 1] asks for the value at the first row and second column. Use the kind of identifier the task specifies, rather than treating positions as labels.
9. How do you select one column versus multiple columns?
df["sales"] selects one column as a Series. df[["sales", "region"]] selects two columns and keeps the result two-dimensional as a DataFrame. This distinction matters when a downstream method expects one object shape rather than the other.
10. How do you filter rows with one condition?
Create a Boolean mask and use it to select matching rows: df[df["sales"] > 100]. The comparison is evaluated across the column, and rows where the condition is true are retained.
11. How do you combine multiple filter conditions?
Put each comparison in parentheses and use element-wise operators such as & for AND and | for OR: df[(df["sales"] > 100) & (df["region"] == "West")]. These operators combine masks element by element; Python’s scalar and and or are not substitutes for this pattern.
12. How do you select rows using an index value?
Use label-based access, for example df.loc["customer_42"], when that value is an index label. If the label occurs more than once, the result can contain multiple rows. Check the index and its uniqueness instead of assuming the label identifies exactly one record.
13. How do you add or derive a column?
Assign a vectorized expression when a calculation applies across a column: df["net"] = df["gross"] - df["discount"]. This expresses the transformation directly over the Series values and their alignment, rather than describing a Python loop over individual rows.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →14. What is reindexing?
Reindexing aligns an object to a requested set or order of labels. For example, series.reindex(["a", "b", "c"]) returns those index labels in that order; a requested label absent from the original index is introduced with a missing value unless a fill behavior is specified. Consider how those new missing values affect later calculations.
Cleaning and missing data
15. How do you detect missing values?
isna() marks missing entries as true, while notna() marks present entries as true. Use these masks to inspect particular fields or summarize missingness before deciding how to handle it. pandas documents detection and treatment options in its missing-data guide.
16. How do you drop rows or columns with missing data?
Use dropna, making the axis and retention rule explicit. For example, df.dropna(axis=0, subset=["email"]) removes rows missing an email value; df.dropna(axis=1, how="all") removes columns that are entirely missing. Choose the rule based on which fields are required and how much data can be excluded.
17. How do you fill missing data?
fillna can insert a constant, such as a documented “unknown” category, or a statistic computed for the relevant population. Forward- or backward-filling propagates neighboring observations in an ordered series. These choices mean different things: a fill value can alter distributions, and propagation is only defensible when the ordering and domain make it meaningful.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →18. What is interpolation, and when might it make sense?
Interpolation estimates values between observed points, using a method suited to the data’s order and meaning. It may be reasonable for a continuous measurement over an ordered time series, but not automatically for categorical labels or observations where intermediate values have no meaningful interpretation. pandas provides interpolation methods as part of its missing-data tools.
19. How do you find duplicate rows?
Use df.duplicated() to identify rows repeated under the chosen comparison and drop_duplicates to remove them if appropriate. Decide which columns define a duplicate and whether the retained record should be the first, last, or a separately resolved authoritative row. Identical-looking records are not necessarily invalid without domain context.
20. How do you replace inconsistent values or labels?
Normalize values to a consistent representation and use replace for known substitutions, such as mapping inconsistent spellings of a region to one label. First identify the variants and confirm they represent the same category; a broad replacement can collapse distinct values accidentally.
21. Why can missing-value treatment change an analysis?
Dropping records changes which observations contribute to later summaries; filling them adds assumed values to those summaries. The choice can therefore change counts, averages, comparisons, and conclusions. Explain which records were affected and why the chosen treatment fits the analysis.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Grouping and aggregation
22. What does groupby do?
GroupBy follows a split-apply-combine pattern: divide observations into groups using keys, apply a calculation within each group, then combine the results. The GroupBy guide describes the operation and its use cases.
Rank #3
- Crisp writing pages are perfect for personal reflections, sketching, or for recording favorite quotations or poems.
- Premium 120 gsm paper takes pen or pencil beautifully.
- Paper is acid free and of archival quality.
- Light gray lines subtly guide your writing.
- An inside back cover pocket expands to hold notes, cards, mementos, and more.
23. How do agg, transform, and filter differ?
agg summarizes each group, often returning fewer rows than the input. transform returns group calculations aligned with the original observations, making it useful when each row needs its group’s value. filter keeps or removes whole groups based on a condition. Choose by the output shape the next step needs.
24. How do you compute several summary measures by group?
Apply multiple aggregations to a grouped object and name what each result means. For example, df.groupby("region").agg(order_count=("order_id", "count"), mean_sales=("sales", "mean")) produces summaries by region. Check whether the count should include only nonmissing order IDs or all rows before selecting the counting method.
25. How do you group by more than one key?
Pass multiple grouping keys, such as df.groupby(["region", "year"])["sales"].sum(). Each combination of region and year defines a group, so the output is organized by those keys, commonly as index levels.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match26. How can you compute a group statistic for every original row?
Use a transformation when each observation needs a result calculated from its group. For instance, df["region_mean"] = df.groupby("region")["sales"].transform("mean") assigns the region’s mean sales to each corresponding row while preserving the row-level shape.
27. How do you count rows or nonmissing values by group?
Group size counts rows in each group, including rows whose measured value is missing. A value count such as df.groupby("region")["sales"].count() counts nonmissing sales values instead. State which quantity answers the question rather than using “count” ambiguously.
28. How do sorting and group output labels affect presentation?
Decide how group keys should be ordered and whether they should appear as index levels or ordinary columns. Set relevant GroupBy options such as sort or as_index deliberately, and inspect the result before passing it to code that expects a particular layout.
Combining data
29. How do merge, join, and concat differ?
merge performs key-based joins similar to SQL joins; join combines objects along columns and commonly uses index alignment; concat combines objects along an axis, such as stacking rows. The merging guide explains these combination patterns.
30. How do you perform an inner, left, right, or outer merge?
Choose the join type according to which unmatched keys must remain. An inner join retains keys matched on both sides; a left join retains all left-side keys; a right join retains all right-side keys; an outer join retains keys from either side. For example, left.merge(right, on="customer_id", how="left") preserves every left-side customer row, with missing right-side fields where no match exists.
31. What causes duplicate rows after a merge?
Non-unique keys on either side can produce multiple matches. If a key appears twice on the left and three times on the right, matching that key can produce six joined rows. Check key uniqueness, expected join cardinality, and row counts; use the merge validate option when a one-to-one, one-to-many, or many-to-one relationship is expected.
32. How do you merge on differently named key columns?
Specify each key explicitly: left.merge(right, left_on="customer_code", right_on="code", how="inner"). This makes clear which fields connect the tables and avoids relying on identical column names.
Rank #4
- Your Everyday Productivity Tool: This wide-ruled notebook offers a reliable space to capture notes, ideas, and plans. Designed for professionals and students who need structure and clarity throughout their busy day.
- Sleek and Durable Design: With a soft faux leather hardcover and strong sewn binding, this compact 5.75" x 8.25" notebook is built to endure daily use, fitting easily into backpacks or briefcases.
- Premium Paper Quality: 120 GSM thick paper resists ink bleed-through and feathering, providing a smooth writing experience for all types of pens and markers.
- Wide Lines for Neat, Comfortable Writing: The wide-ruled format allows you to write clearly and comfortably, reducing hand strain and making it easy to stay organized during lectures, meetings, or journaling.
- Versatile Notebook for All Needs: Whether you’re managing work tasks, school notes, or personal projects, this notebook helps keep everything in one place for easy access and productivity.
33. How do you combine DataFrames stacked vertically?
Use pd.concat([jan, feb], axis=0) to append rows. By default, the existing indexes are carried into the result, so consider whether to preserve them or use ignore_index=True for a new sequential index. Confirm that similarly named columns mean the same thing across inputs.
Free tools Windows power users keep installed
One-click scans. No signup required.
34. How do you join using indexes?
Use index-based joining when the row indexes are the intended keys, for example left.join(right, how="left"). That differs from a merge that names key columns explicitly. Verify the indexes represent the same entity or time grain before aligning them.
35. How can you diagnose unmatched keys?
Use an outer merge with an indicator to label rows as matched or present only on one side, then inspect the unmatched cases. Alternatively, compare key sets in each direction. Check whitespace, spelling, type differences, and missing keys before concluding that records are genuinely absent; confirm option support against the pandas version in use.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Reshaping
36. What does it mean to reshape wide data into long data?
Wide data often has one column per measurement or time period; long data stores measurements in rows, with identifier columns describing the entity and separate columns for the measurement name and value. melt is a common way to move selected measurement columns into this longer form, which can make grouping and plotting more straightforward.
37. What do pivot and pivot_table solve?
pivot rearranges values when each index-and-column combination identifies one value. Repeated combinations make that mapping ambiguous. pivot_table handles repeated combinations by aggregating them, using a chosen aggregation function. Decide whether aggregation is valid for the data rather than using it merely to conceal duplicates.
Recommended Free Tools
38. What do stack and unstack do?
These operations move levels between columns and the row index. In broad terms, stacking moves column levels into the index, while unstacking moves an index level into columns. They are useful when changing the layout of data organized by multiple labels.
39. How do you remove duplicate observations before reshaping?
Check whether the intended index-and-column keys are unique before a one-value-per-cell pivot. If duplicate observations exist, determine whether they are erroneous duplicates or legitimate repeated measurements. Remove only confirmed duplicates; otherwise choose an aggregation that reflects the analysis.
40. How do you choose a useful output layout?
Choose a shape that fits the next task. Long layouts often make categories and measurements easier to group, chart, or join; wide layouts may be clearer for a report or a model expecting separate features. Keep identifier fields explicit and avoid reshaping away distinctions the analysis needs.
Time series
41. How do you parse strings as dates when reading a dataset?
Use date-parsing options when reading a file or convert the column afterward with pd.to_datetime. Check the resulting dtype and inspect invalid or ambiguous entries. Parsing is not complete just because values look date-like when printed.
42. What is a datetime index useful for?
A datetime index supports time-based selection and operations such as resampling. It lets code refer to dates or time ranges as labels, provided the values have been parsed correctly and the intended time zone is understood.
43. What is resampling?
Resampling changes the frequency of time-indexed data by assigning observations to time bins and applying an aggregation or fill operation. For example, daily measurements might be summarized by month. Choose a frequency and aggregation that match what each observation represents.
44. How do rolling windows differ from calendar resampling?
A rolling calculation evaluates a moving window around observations, such as a moving average over the previous observations. Resampling groups timestamps into calendar or other defined bins and summarizes each bin. One moves across observations; the other forms time buckets.
45. How should time zones be handled?
Distinguish localization from conversion. Localization assigns a time zone to naive timestamps whose local meaning is known; conversion changes already timezone-aware timestamps to another zone while preserving the instant. Confirm the source zone and the intended reporting zone, especially around daylight-saving transitions.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteInput, output, and scale
46. How do you read a CSV file?
Use pd.read_csv("file.csv"), then inspect the parsed columns and types. Options such as usecols can limit the fields read, while dtype or date-parsing choices can help represent the data appropriately. Validate the result rather than assuming the file’s headers and inferred types are correct.
47. How can you process a CSV in chunks?
Pass chunksize to read_csv to receive an iterator of DataFrame chunks, then process each chunk and combine only the summaries or results needed. For example, a count or sum can be accumulated incrementally when the calculation permits it. pandas documents chunk and iterator options in its IO guide.
48. How do you write a DataFrame to a file?
Choose an export method that matches the recipient and data use, such as to_csv for a CSV file. Decide explicitly whether to include the index: a row index that is meaningful to the recipient may be part of the output, while an incidental index often should be omitted with index=False.
49. What are reasonable first steps when pandas code is slow?
Identify the slow operation and measure it before changing the implementation. Check whether the workflow reads or processes unnecessary rows and columns, and avoid unnecessary Python-level work one row at a time when a vectorized operation can express the calculation. Compare timings and verify that any change preserves the result.
50. When might data exceed a single in-memory DataFrame workflow?
If the dataset or intermediate results cannot be handled comfortably in memory, consider incremental processing with chunks where the computation allows it. If the task requires repeated cross-chunk operations or data volumes remain impractical, a storage or processing system designed for that workload may be more suitable. The right boundary depends on the machine, data, and operation rather than one universal row-count threshold.
51. How do you explain a pandas solution in a live interview?
State the assumptions first: what defines a record, which fields are keys, how missing values should be treated, and whether keys are unique. Then describe the transformations in order, explain the expected output shape and labels, and check row counts or key cardinality where joins or filtering could change them. This makes the solution’s logic and data-quality dependencies clear.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
