Key takeaways
- MCQs test recall and recognition, so pair them with coding practice.
- Read each explanation, including for questions you answered correctly.
- Group mistakes by topic and revise the weakest topic first.
- Timed practice builds exam speed.
- Distractors usually reflect real misunderstandings, such as confusing correlation with causation.
What is data science multiple choice questions?
How do I use MCQs well?
Attempt a set without notes, mark the ones you guessed, then read the answers. For each miss, write the idea in one line. Repeat the set after a few days. Practice with explanations teaches more than a score.
Statistics practice
Q1. Which measure is least affected by outliers? A) Mean B) Median C) Range D) Variance. Answer: B. The median depends on order, not on extreme values.
Q2. A correlation of -0.9 means: A) No relation B) Weak positive C) Strong negative linear relation D) Causation. Answer: C. The sign gives direction and the size gives strength; it does not prove cause.
Q3. A small p-value suggests: A) The null hypothesis is certainly false B) Data would be unlikely if the null were true C) The effect is large D) The sample is biased. Answer: B. It is about the data given the null, and it says nothing about effect size.
Python and pandas practice
Q4. Which call shows missing values per column? A) df.isna().sum() B) df.null() C) df.count_na() D) df.missing(). Answer: A.
Q5. In pandas, which joins keep every row from the left table? A) inner B) left C) outer only D) cross. Answer: B.
Q6. What does df.groupby("city")["sales"].mean() return? A) One row per row B) Average sales per city C) Total sales D) Number of cities. Answer: B.
Machine learning practice
Q7. Overfitting means the model: A) Performs well on training data but poorly on new data B) Performs poorly everywhere C) Is too simple D) Has no features. Answer: A.
Q8. For a rare disease screening test, which metric is most important to watch? A) Accuracy only B) Recall C) Mean error D) Variance. Answer: B. Missing real cases is costly, though precision should be checked too.
Q9. Which algorithm groups unlabeled data? A) Logistic regression B) K-means C) Linear regression D) Naive Bayes. Answer: B.
Q10. Why scale features for k-nearest neighbours? A) To add rows B) Because distance depends on feature size C) To remove labels D) To increase accuracy always. Answer: B.
Evaluation and leakage practice
Q11. Data leakage happens when: A) Test information influences training B) The data file is large C) The model is saved D) A chart is wrong. Answer: A. Fitting a scaler on all data before splitting is a common cause.
Q12. Cross-validation is used to: A) Remove outliers B) Estimate performance more reliably C) Create new labels D) Speed up training. Answer: B.
How do I review my results?
Count correct answers by topic. Anything below your target means returning to notes and coding examples. Create ten of your own questions to teach the idea back.
What question styles appear in written tests?
Written tests mix direct definition questions, scenario questions and code-reading questions. A scenario question may describe a model with high training accuracy and low test accuracy and ask for the likely cause. A code-reading question may show three lines of pandas and ask what the output is. To prepare, practise saying why each wrong option is wrong, since that exposes shallow understanding. Also practise reading the question twice, because words like "not", "least" and "except" change the answer entirely.
How can I make my own practice questions?
After each topic, write five questions with four options each, making the wrong options from real mistakes you or classmates have made. Swap sets with a friend and explain each answer aloud. This method turns passive reading into active recall and gives you a personal bank that targets your weak areas. Keep a running list of mistakes, review it weekly and retire questions you answer correctly three times in a row so that your practice time stays focused where it matters.
How do I manage time in a timed test?
Do a first pass answering only the questions you are sure about, mark doubtful ones and return to them. Do not spend more than a minute on a single item in the first pass. In the second pass eliminate clearly wrong options before choosing. Leave a few minutes at the end to check marked answers and any rule about negative marking in your exam.
Which resources help beyond MCQs?
Work through the official scikit-learn and pandas tutorials, solve small coding exercises and explain concepts aloud. Interviewers often follow a multiple-choice style fact with a "why" question, so understanding the reason behind each answer is more valuable than memorising options. Combine quizzes with a project so ideas connect to real data.
Download Data Science MCQ PDF
Data Science MCQ PDF
The PDF carries the same content as this page in a clean, print-ready layout, so you can read it offline and keep it for revision.
4 pages · 93 KB · PDF · Updated 5 Nov 2026