Statistics Interview Questions for Data Analysts: 23 With Answers

Last updated: 13 October 2026 · By the Asuraa Team

Quick answer: Statistics interviews for data analysts test descriptive statistics, probability, distributions, hypothesis testing and correlation. The most common traps are the median vs mean choice for skewed data and the p-value definition: per the NIST/SEMATECH e-Handbook, a p-value is the probability of a result at least as extreme as the one observed if the null hypothesis is true, and α = 0.05 rejects a true null 5% of the time.

Key takeaways

• The NIST/SEMATECH e-Handbook defines a p-value as the probability of a test statistic at least as extreme as the one observed, assuming the null hypothesis is true.
• A significance level of 0.05 means a true null hypothesis is rejected 5% of the time, and NIST lists 0.1, 0.05 and 0.01 as common choices.
• NIST notes that the median is a better estimate of location than the mean for data with outliers or heavy tails, because it is based on ranks.
• With a 1% base rate, 90% sensitivity and a 5% false-positive rate, only about 15% of positive test results are true positives, which is the classic Bayes interview example.
• Statistical significance shows an effect is unlikely to be chance, but analysts should report effect size and a confidence interval to show whether it matters in practice.

Statistics rounds for fresher data analyst and data scientist roles in India test whether you can summarise data honestly, reason about probability and judge whether a difference is real. Most questions sit at the level of a first statistics course, but interviewers expect clear explanations with an example, not textbook definitions. Below are 23 questions with model answers, formulas and a short Python snippet, a weighting table and a 7-day plan. For other subjects, see the interview questions by subject guide.

Which statistics topics carry the most weight in analyst interviews?

Descriptive statistics and hypothesis testing carry the most weight; probability puzzles and regression are common follow-ups, especially for data scientist roles.

What descriptive statistics questions are asked first?

These are the warm-up questions, and a weak answer here ends the round early.

1. When would you use the median instead of the mean?

Use the median when data is skewed or has outliers, because the mean is pulled towards extreme values. The NIST/SEMATECH e-Handbook notes that for data with heavy tails or outliers, the median is a better estimate of location than the mean, since it depends on ranks. Salaries are the classic example: a few CXO salaries of Rs 2 crore can lift the mean far above what a typical employee earns, while the median stays representative.

2. What is standard deviation, and why does the sample formula divide by n−1?

Standard deviation measures how spread out values are around the mean; it is the square root of variance, the average squared deviation. For a sample, dividing by n−1 instead of n (Bessel's correction) corrects the tendency of a sample to underestimate the population's spread, because the sample mean is itself estimated from the same data.

s = sqrt( Σ (xᵢ − x̄)² / (n − 1) )

3. What is the interquartile range, and how do you use it to find outliers?

The IQR is Q3 minus Q1, the spread of the middle 50% of the data, so it resists outliers. A common rule flags values below Q1 − 1.5×IQR or above Q3 + 1.5×IQR as potential outliers, which is what a box plot's whiskers show. Treat flagged points as values to investigate, not automatically delete.

4. What does skewness tell you?

Skewness describes asymmetry. In a right-skewed (positively skewed) distribution, such as income or delivery times, the long tail is on the right and the mean is usually greater than the median. In a left-skewed distribution, such as exam scores when most students do well, the tail is on the left. Reporting the median with the mean warns readers about skew.

5. What are percentiles and quartiles?

The pth percentile is the value below which p% of observations fall; quartiles are the 25th, 50th (median) and 75th percentiles. They are useful for service metrics: saying "90% of orders were delivered within 38 minutes" is more informative than an average. Percentiles also help compare scores across groups of different sizes.

What probability questions should you prepare?

Interviewers want to see your working, so say each step aloud.

6. What is conditional probability, and how does Bayes' theorem use it?

Conditional probability P(A|B) is the probability of A given that B happened, equal to P(A and B) / P(B). Bayes' theorem reverses the condition: P(A|B) = P(B|A) × P(A) / P(B). Example: a disease affects 1% of people, a test catches 90% of cases and gives 5% false positives. P(positive) = 0.9×0.01 + 0.05×0.99 = 0.0585, so P(disease | positive) = 0.009 / 0.0585 ≈ 15%. Low base rates make most positives false.

7. What is the difference between independent and mutually exclusive events?

Independent events do not affect each other's probability, so P(A and B) = P(A) × P(B), like two coin tosses. Mutually exclusive events cannot happen together, so P(A and B) = 0, like a single die showing both 2 and 5. Two events with non-zero probability cannot be both, because if one happening rules out the other, they clearly affect each other.

Which distribution questions come up?

Know four distributions and when each fits.

8. What is the normal distribution, and what is the 68-95-99.7 rule?

The normal distribution is the symmetric bell curve defined by its mean and standard deviation, where mean, median and mode coincide. About 68% of values lie within one standard deviation of the mean, about 95% within two and about 99.7% within three. Heights and measurement errors are roughly normal; incomes and website session lengths usually are not.

9. What is the central limit theorem, and why does it matter?

The central limit theorem says that the distribution of sample means approaches a normal distribution as sample size grows, whatever the shape of the original data, provided observations are independent with finite variance. It matters because it lets us build confidence intervals and run t-tests on means even when individual values are skewed, such as order values. The standard error of the mean is σ/√n.

10. When do you use the binomial and Poisson distributions?

The binomial models the number of successes in a fixed number of independent trials with the same probability, such as how many of 100 visitors click a button with a 3% click rate. The Poisson models the count of events in a fixed interval when events occur independently at a constant average rate, such as calls arriving at a support centre per minute. Binomial has a fixed upper limit; Poisson does not.

What hypothesis testing questions are asked?

This is the core of most statistics rounds; expect at least three questions here.

11. What are the null and alternative hypotheses?

The null hypothesis (H0) is the default claim of no effect or no difference, such as "the new checkout page has the same conversion rate as the old one". The alternative (H1) is what you are trying to find evidence for, such as "the new page's rate is different". A test either rejects H0 or fails to reject it; we never "accept" H0, because absence of evidence is not proof of no effect.

12. What is a p-value?

The NIST e-Handbook on significance tests defines the p-value as the probability of getting a test statistic at least as extreme as the one observed, assuming the null hypothesis is true. A small p-value means the data would be unusual if H0 were true, so you reject H0 when p is below the significance level you fixed in advance. It is not the probability that H0 is true.

13. What are Type I and Type II errors, and what is the significance level?

A Type I error is rejecting a true null hypothesis (a false positive); a Type II error is failing to reject a false one (a false negative). The significance level α is the Type I error rate you accept; NIST explains that α = 0.05 means a true null hypothesis is rejected 5% of the time, and lists 0.1, 0.05 and 0.01 as common choices. Power, 1 − β, is the probability of detecting a real effect, and larger samples increase it.

14. How do you interpret a 95% confidence interval?

If you repeated the sampling many times and built an interval each time, about 95% of those intervals would contain the true population value. It does not mean there is a 95% probability that this particular interval contains it. A narrower interval, from a larger sample or less variable data, means a more precise estimate.

15. When do you use a t-test, and what types are there?

Use a t-test to compare means when the population standard deviation is unknown, which is almost always. A one-sample t-test compares a mean to a fixed value, an independent two-sample t-test compares two separate groups (Welch's version does not assume equal variances), and a paired t-test compares before-and-after measurements on the same units. The z-test is used when the population standard deviation is known or for proportions with large samples.

from scipy import stats
t_stat, p_value = stats.ttest_ind(group_a, group_b, equal_var=False)  # Welch's t-test
print(round(p_value, 4))

16. When do you use a chi-square test?

Use the chi-square test of independence to check whether two categorical variables are related, such as city and preferred payment method, using a contingency table of counts. It compares observed counts with the counts expected if the variables were independent. Expected counts in each cell should not be too small; otherwise use Fisher's exact test.

17. What is ANOVA?

One-way ANOVA tests whether the means of three or more groups differ, such as average order value across four cities, without inflating the Type I error that many pairwise t-tests would cause. A significant result says at least one mean differs, not which one, so follow it with a post-hoc test such as Tukey's HSD.

What correlation and regression questions come up?

These check that you do not over-claim from data.

18. What is the difference between correlation and causation?

Correlation measures how strongly two variables move together; causation means one directly changes the other. Ice cream sales and drowning cases both rise in summer because of a third factor, hot weather, not because one causes the other. To argue causation you need a controlled experiment such as an A/B test, or careful methods that account for confounders.

19. What is the difference between Pearson and Spearman correlation?

Pearson's r measures the strength of a linear relationship between two numeric variables and is sensitive to outliers. Spearman's rho measures a monotonic relationship using ranks, so it handles outliers and non-linear but consistently increasing relationships better. Both range from −1 to +1, and a value near 0 means no relationship of that type, not no relationship at all.

20. What does R-squared mean in linear regression?

R-squared is the share of variance in the target explained by the model, from 0 to 1. An R-squared of 0.65 for predicting house prices from area means area explains 65% of price variation in that data. A high R-squared does not prove causation or a good model; check residual plots, and use adjusted R-squared when comparing models with different numbers of predictors.

What applied statistics questions are asked?

These scenario questions separate analysts who can use statistics from those who can only define it.

21. How would you design an A/B test?

Define one primary metric, such as checkout conversion, and the minimum change worth detecting. Calculate the sample size needed for your chosen α and power, randomly split users into control and variant, and run the test for the full planned duration, ideally covering whole weeks. Do not stop early when the p-value first dips below 0.05; repeated peeking inflates false positives.

22. How do you handle missing data?

First find out why it is missing: completely at random, related to other variables, or related to the missing value itself. If few rows are affected and missingness is random, dropping them is acceptable. Otherwise impute with the median for skewed numbers, the mode for categories, or a model-based method, and add a flag column so the analysis knows which values were filled.

23. What is the difference between statistical and practical significance?

Statistical significance means an effect is unlikely to be due to chance; practical significance means it is large enough to matter. With a million users, a 0.01 percentage point lift in conversion can be statistically significant yet not worth the engineering cost. Report the effect size and confidence interval alongside the p-value so decision-makers can judge.

What do most guides on statistics interview questions get wrong?

Most guides define the p-value as "the probability that the null hypothesis is true", which is wrong. NIST's definition is the probability of a result at least as extreme as yours if the null is true; it says nothing directly about whether H0 is true. Guides also present 0.05 as a law, while NIST treats the choice of α as somewhat arbitrary and stresses fixing it before you test. A third common error is calling a 95% confidence interval a 95% chance that the true value is inside this one interval. Getting these three right in an interview marks you out quickly.

How should you revise statistics in 7 days?

Pair every concept with a small calculation in Python or Excel on real data.

1. Day 1: Mean, median, mode, standard deviation, IQR and skew on a salary or delivery-time dataset.
2. Day 2: Probability rules, independence, Bayes' theorem; solve five word problems aloud.
3. Day 3: Normal, binomial and Poisson distributions; simulate the central limit theorem in Python.
4. Day 4: Hypothesis tests: H0 and H1, p-values, α, Type I and II errors, power.
5. Day 5: t-tests, chi-square and ANOVA with scipy on a public dataset.
6. Day 6: Correlation, regression, R-squared; design an A/B test with a sample-size calculation.
7. Day 7: Answer all 23 questions aloud in under 90 seconds each and explain one analysis from your project.

For the underlying concepts, read statistics and probability for data science. Then cover the rest of the round with the data analyst interview questions and data scientist interview questions, and use how to prepare for technical interview questions for the round format.

Related guides

• Interview Questions by Subject: Java, Python, SQL, DSA, Cloud, AI
• Power BI Interview Questions (2026): 22 Questions With Answers
• AWS Interview Questions for Freshers: 22 Questions With Answers
• Excel Interview Questions for Data Analysts: 23 Questions With Answers

FAQ

What statistics should a data analyst know for interviews?

Focus on descriptive statistics (mean, median, standard deviation, percentiles, IQR), basic probability and Bayes' theorem, the normal, binomial and Poisson distributions, the central limit theorem, hypothesis testing with p-values and Type I and II errors, t-tests and chi-square tests, correlation and simple regression, and how to design an A/B test. Practise explaining each with a business example.

How do you explain a p-value in simple words?

Say: if there were truly no effect, the p-value is the chance of seeing a result at least as extreme as ours. A small p-value means our result would be surprising under 'no effect', so we reject that assumption when p falls below the level we set beforehand, often 0.05. It is not the probability that the null hypothesis is true.

What is the difference between Type I and Type II errors?

A Type I error is a false positive: rejecting a null hypothesis that is actually true, such as concluding a new page lifts sales when it does not. A Type II error is a false negative: missing a real effect. The significance level controls the Type I rate, and larger samples raise power, which lowers the Type II rate.

Is statistics asked in data analyst interviews in India?

Yes, usually as part of the technical round alongside SQL and Excel. Fresher analyst interviews in India commonly ask about mean vs median, standard deviation, outliers, correlation vs causation and basic hypothesis testing. Data scientist roles go further into distributions, regression assumptions and experiment design. Expect at least a few questions even when the role is SQL-heavy.

When should I use median instead of mean?

Use the median when data is skewed or contains outliers, such as salaries, house prices or delivery times. The mean is pulled towards extreme values, while the median depends only on the middle of the ordered data. Reporting both, along with the spread, gives readers the clearest picture of a skewed distribution.

What is the central limit theorem in simple terms?

If you take many random samples of a reasonable size and compute each sample's mean, those means form a roughly normal distribution around the true mean, even if the original data is skewed. This is why analysts can build confidence intervals and run t-tests on averages like order value. The spread of those means shrinks as sample size grows.

Final thoughts

Statistics interviews reward careful explanation more than formulas: say what a number means, what could mislead, and what you would check next. Work through the 7-day plan on a real dataset, and if you want feedback on how you explain an analysis, a session with an Asuraa mentor can help.