Statistics & Experimentation Interview Questions

Farza S. - "It is the smallest level of significance at which the null hypothesis gets rejected"See full answer

Sangeeta P. - "A/B testing is used when one wishes to only test minor front-end changes on the website. Consider a scenario where an organization wishes to make significant changes to its existing page, such as wants to create an entirely new version of an existing web page URL and wants to analyze which one performs better. Obviously, the organization will not be willing to touch the existing web page design for comparison purposes. In the above scenario, performing Split URL testing would be beneficial. T"See full answer

Lucas G. - "The central limit theorem tells us that as we repeat the sampling process of an statistic (n > 30), the sampling distribution of that statistic approximates the normal distribution regardless of the original population's distribution. This theorem is useful because it allows us to apply inference with tools that assume normality like t-test, ANOVA, calculate p-values hypothesis testing or regression analysis, calculate confidence intervals, etc."See full answer

Lucas G. - "I would conduct a sample z-test because we have enough samples and the population variance is known. H1: average monthly spending per user is $50 H0: average monthly spending per user is greater $50 One-sample z-test x_bar = $85 mu = $50 s = $20 n = 100 x_bar - mu / (s / sqrt(n) = 17.5 17.5 is the z-score that we will need to associate with its corresponding p-value. However, the z-score is very high, so the p-value will be very close to zero, which is much less than the standa"See full answer

Mark S. - "Range captures the difference between the highest and lowest value in a data set, while standard deviation measures the variation of elements from the mean. Range is extremely sensitive to outliers, it tells us almost nothing about the distribution of the data, and does not extrapolate to new data (a new value outside the range would invalidate the calculation). Standard deviation, on the other hand, offers us an insight into how closely data is distributed towards the mean, and gives us some pr"See full answer

Lucas G. - "To speed up A/B tests results with limited sample sizes, we can apply advanced techniques like CUPED to reduce variance for faster statistical significance, interleaving to gather more comparative data per user (e.g., ranking), MAB to dynamically allocate traffic to winning variations for quicker optimization (e.g., campaigns), and Bayesian A/B testing which offers probabilistic conclusions that can be reached earlier. Each method, when appropriately applied, allows you to gain m"See full answer

Lucas G. - "Because testing many engagement metrics at once increases the risk of finding effects that aren't real (the 'multiple comparisons problem'), you must adjust your criteria for statistical significance. For social media data, the Benjamini-Hochberg procedure is often a practical choice as it controls the rate of false discoveries (FDR) while still allowing you to detect genuine changes; however, the ideal adjustment method will vary depending on your specific number of metrics (e.g., use Bonferron"See full answer

Connor W. - "This video is a duplicate of the other video in this lesson, "Design A/B test for New Campaign""See full answer

Saurabh K. - "Ask a follow up question : What is the primary goal of expanding into a new vertical ? Food vertical company may want to expand to a new vertical (say Grocery) for the following reasons : Attract new customers interested in grocery delivery instead of food delivery Increase usage/order frequency from existing customers Increase revenue and LTV of existing as well as potentially new customers Benefit from synergies between existing delivery engine by improving utilization of their network"See full answer

Sangeeta P. - "Marketing campaigns are run through different channels such as social media, emails, SEO, web advertising, events, etc. Let’s look at some of the overall success metrics at a broader level: Total views for your campaign Unique views for your campaign Returning visitors for your campaign Engagement for your campaign (If it’s a social media campaign, the marketer might be interested in knowing the number of users engaging with the campaign and the type of campaign positive/negative) 5"See full answer

Himani E. - "The cases where data is under heavy outlier influence. Since mean fluctuates due to the presence of an outlier, median might be a better measure"See full answer

Lucas G. - "Null hypothesis (H0): the coin is fair (unbiased), meaning the probability of flipping a head is 0.5 Alternative (H1): the coin is unfair (biased), meaning the probability of flipping a head is not 0.5 To test this hypothesis, I would calculate a p-value which is the probability of observing a result as extreme as, or more extreme than, what I say in my sample, assuming the null hypothesis is true. I could use the probability mass function of a binomial random variable to model the coin toss b"See full answer

Emiliano I. - "Look for the main variables and see if there differences in the distributions of the buckets. Run a linear regression where the dependent variable is a binary variable for each bucket excluding one and the dependent variable is the main kpi you want to measure, if one of those coefficients is significant, you made a mistake. "See full answer

Lucas G. - "A confidence interval gives you a range of values where you can be reasonably sure the true value of something lies. It helps us understand the uncertainty around an estimate we've measured from a sample of data. Typically, confidence intervals are set at the 95% confidence level. For example, A/B test results show that variant B has a CTR of 10.5% and its confidence intervals are [9.8%, 11.2%], this means that based on our sampled data, we are 95% confident that the true avg CTR for variant B a"See full answer

Lucas G. - "Type I error (typically denoted by alpha) is the probability of mistakenly rejecting a true null hypothesis (i.e., We conclude that something significant is happening when there's nothing going on). Type II (typically denoted by beta) error is the probability of failing to reject a false null hypothesis (i.e., we conclude that there's nothing going on when there is something significant happening). The difference is that type I error is a false positive and type II error is a false negative. T"See full answer