Skill readiness test — Senior Data Scientist

10 questions, about 6 minutes. This is a diagnostic, not a certification — it estimates your skill readiness, it does not guarantee you are qualified for or will be hired into this role.

1. What does `df.groupby('region')['sales'].sum()` return?

The total sales across the whole dataframe Total sales summed for each distinct region A count of rows per region An error, because groupby requires two columns

2. Which measure of central tendency is least affected by extreme outliers?

Mean Median Range Standard deviation

3. A query aggregates revenue by region but SELECT also lists an ungrouped column. What happens on strict SQL engines (e.g. MySQL with ONLY_FULL_GROUP_BY, or PostgreSQL)?

It runs and picks an arbitrary value for the ungrouped column It errors — every selected column must be aggregated or in GROUP BY It silently returns NULL for the ungrouped column It automatically adds the column to GROUP BY

4. In Power BI, what is the main purpose of a DAX measure versus a calculated column?

Measures compute dynamically per visual context; calculated columns are stored per row They are identical in every way Calculated columns can only be text Measures cannot use aggregation functions

5. Two tables are related on a shared key but the relationship shows a dotted line in the model view. What does that usually mean?

The relationship is inactive The tables are identical The relationship is many-to-many by definition There is a syntax error

6. Which library is most commonly used for tabular data manipulation in Python?

pandas flask requests pytest

7. Which clause filters rows before they are grouped by GROUP BY?

HAVING WHERE ORDER BY LIMIT

8. You need each customer's most recent order. Which approach is most reliable?

ORDER BY order_date DESC LIMIT 1 with no grouping A window function (ROW_NUMBER() OVER (PARTITION BY customer_id ORDER BY order_date DESC)) MAX(order_date) alone with SELECT * GROUP BY customer_id with no aggregate on the other columns

9. A p-value of 0.03 from a hypothesis test at a 0.05 significance threshold suggests:

The null hypothesis is definitely false There is statistically significant evidence against the null hypothesis The result is not usable The sample size was too small

10. An INNER JOIN between orders and customers drops rows where...

the customer has no matching order the order has no matching customer both a and b nothing is dropped

Senior Data Scientist at Starbucks