AMAADOR ACADEMYCertificatesVerify a certificate

STATX

Final check — Statistics and A/B Testing for Data Analysts

Answer the 42 questions. You need 80% or more to pass. You can retry as many times as you like.

Study the course →

Question 1Which measure of centre is the least affected by one extreme value?
Question 2Five order values in euros are 20, 22, 25, 27 and 400. The mean is 98.8. You must report the typical order value. What is the best choice?
Question 3A dataset has a column "plan" with values Basic, Pro and Team. An analyst codes them 1, 2 and 3 and reports a mean plan of 1.8. What is the problem?
Question 4What does the interquartile range (IQR) measure?
Question 5A box plot of delivery times shows six points far above the upper whisker. The monthly report is due in an hour. What should you do first?
Question 6Two support teams both average 10 tickets a day. Team A is always between 9 and 11. Team B ranges from 0 to 25. Which statistic shows the difference between them?
Question 7What does the standard error of a mean describe?
Question 8Session lengths have a standard deviation of 20 minutes. You sample 100 sessions. What is the standard error of the mean?
Question 9Your estimate of average basket value has a standard error of 4 euros. The director wants it cut to 2 euros. How much data do you need compared with now?
Question 10Which problem does NOT get smaller when you collect more data from the same biased source?
Question 11A retention report for a game uses only players who are still active on day 30 and concludes that "players love the tutorial". What is the main risk?
Question 12An opt-in poll on your website gets a 3% response rate. The people who reply are mostly very happy or very angry customers. What is the problem with treating the result as the view of all customers?
Question 13Which statement about a 95% confidence interval is correct?
Question 1452 of 400 visitors signed up (13%). The standard error of that proportion is about 1.7 points. Which is the approximate 95% confidence interval?
Question 15A new feature lifts a metric by 0.8 points, with a 95% confidence interval from -0.3 to +1.9 points. What can you say?
Question 16You recompute the same interval at 99% confidence instead of 95%, using the same data. What happens to its width?
Question 17When is the bootstrap an appropriate way to get a confidence interval?
Question 18You measured 12 call durations and need a 95% interval for the mean. A colleague multiplies the standard error by 1.96. What is the better approach?
Question 19Which sentence defines a p-value correctly?
Question 20A colleague says: "p = 0.03 means there is a 97% chance that variant B is truly better." How should you respond?
Question 21You compare average handling time between two independent groups of agents. The groups have clearly different variances and sizes. Which test is the safer default?
Question 22You measure the same 40 employees’ task times before and after a training course. Which analysis fits best?
Question 23What is a Type I error?
Question 24A test on 60 users gives p = 0.20 and a very wide interval. A manager writes: "The test proves the redesign has no effect." What is the best correction?
Question 25Two variables have a Pearson correlation of r = 0.9. What does this tell you?
Question 26Monthly ice-cream sales and drowning accidents are strongly correlated. A journalist writes that ice cream causes drowning. What is the most likely explanation?
Question 27A university admits 44% of male and 35% of female applicants overall. Department by department, women are admitted at similar or higher rates. What explains this pattern?
Question 28A regression of weekly sales on ad spend (both in thousands) gives a slope of 3.0 with a 95% interval of 2.1 to 3.9. How do you read the slope?
Question 29A simple regression has an R squared of 0.64. A colleague says "the model is 64% accurate". What is the correct reading?
Question 30Your regression was fitted on ad spend between 1,000 and 10,000. A manager asks what sales would be at 100,000 of spend. What should you say?
Question 31Why does an A/B test assign users to groups at random?
Question 32A team randomises each page view to version A or B. The same visitor often sees both versions during one visit. Which is the better design?
Question 33Baseline sign-up is 5%. You want 80% power at alpha 0.05 to detect a lift to 6%. Roughly how many users per variant do you need?
Question 34Your sample size plan targets a 1-point minimum detectable effect. Leadership now says a 0.5-point lift would also matter. How does the required sample change?
Question 35A new checkout raises conversion, but median page load time rises by 40% in the variant. In the test plan, what role does load time play?
Question 36A 50/50 test records 10,000 users in A and 10,500 in B. A chi-square check gives p of about 0.0005. What does this indicate?
Question 37You run 20 independent tests at alpha 0.05 when no real effect exists anywhere. What is the chance of at least one false positive?
Question 38A homepage test shows no overall lift. Checking 12 segments, an analyst finds p = 0.03 for one of them and proposes to announce it. What is the right handling?
Question 39A fixed-horizon test is checked every morning, and the team stops when p first falls below 0.05. What is the main statistical consequence?
Question 40With 5 million users, a feature lifts a metric by 0.02 points with p < 0.001. Shipping costs three engineer-months. What is the sound conclusion?
Question 41The smallest worthwhile lift is 0.5 points. Your 95% interval for the lift is -0.2 to +2.2 points. Which recommendation is most honest?
Question 42The primary metric shows no clear lift, but one secondary metric improved with p = 0.04. You are writing the summary for leadership. Which line is the most honest?