Statistics · Module 5Module material
Estimation interview questions
25 questions interviewers ask on this module’s topics, from first principles to the follow-ups. Each comes with a model answer.
Sampling distributions
- What is a sampling distribution?●
- What is a standard error, and how does it differ from a standard deviation?●
- Why does the standard error shrink as n grows, and what does the finite population correction add?●●
- What does the central limit theorem say, and what does it not say?●●
- Why is "n = 30 is enough" a poor rule?●
Maximum likelihood
- What is a likelihood?●
- Why maximize the log likelihood instead of the likelihood?●
- Derive the Bernoulli MLE.●●
- How is log loss related to likelihood?●●
- A feature lowers in-sample log loss. Does it improve prediction?●●
Confidence intervals
- What does a 95% confidence interval mean?●
- Why use a Wilson interval for a proportion rather than the Wald interval?●●
- How would you use an interval for a threshold decision?●●
- How can you check whether an interval procedure works for your data?●●
- Why does a t interval undercover on a heavy-tailed frame?●●
The bootstrap
- What is the bootstrap?●
- Why must bootstrap resampling use replacement?●
- What is a paired bootstrap, and when is it needed?●●
- Why do ties matter for ROC-AUC?●●
- What does a bootstrap interval over fixed fitted models leave out?●●
Bayesian estimation
- What is a prior, and what is a posterior?●
- Why is the Beta distribution convenient for a Bernoulli rate?●
- Why does a posterior depend so much on the prior when data are few?●●
- How does a credible interval differ from a confidence interval?●●
- What must hold before a posterior rate from an archive describes the future?●●