← Google Interview Insights

Google·Data Scientist·Technical Phone Screen·Senior

SeniorPrefer not to say
Jul 2026Remote

Summary

Technical phone screen for a Data Scientist role at Google. One statistics question that felt deceptively simple but had a real gotcha buried in it.

Questions Asked (1)

Q1

Can bootstrap help reduce variance?

Technical Trade-offsAlgorithms & Data Structures
Author's notes

I said yes immediately and started talking about how bootstrap gives you a variance estimate, which...

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Clarify that the question is about using bootstrap resampling to estimate the variance of a statistic, and explain that bootstrap provides a direct, non-parametric estimate of variance without relying on distributional assumptions. Then discuss when bootstrap reduces variance compared to other methods, such as when analytical formulas are unavailable or when the sample size is small, and mention potential pitfalls like bias and computational cost.

Pro tip: Emphasize that bootstrap estimates the variance of a statistic, not the variance of the data itself, and that it can actually increase variance if the resampling scheme is not appropriate (e.g., for dependent data).

1. Define the goal

State that the question asks whether bootstrap can reduce variance, and clarify that 'reduce variance' means obtaining a more accurate or lower-variance estimate of a statistic's sampling variance.

2. Explain bootstrap basics

Briefly describe bootstrap: resampling with replacement from the observed data to approximate the sampling distribution of a statistic, and using the empirical variance of the bootstrap replicates as an estimate of the statistic's variance.

3. Compare to analytical methods

Discuss scenarios where bootstrap reduces variance relative to analytical approximations, such as when the analytical formula is biased or when the statistic is complex (e.g., median, correlation).

4. Discuss limitations and trade-offs

Mention that bootstrap does not always reduce variance; it can be biased for small samples, computationally intensive, and inappropriate for dependent data without modifications (e.g., block bootstrap).

5. Conclude with practical implications

Summarize that bootstrap is a powerful tool for variance estimation when used appropriately, but it is not a universal variance reducer; its effectiveness depends on the context and implementation.

Key Points to Mention

  • Bootstrap estimates the sampling distribution of a statistic by resampling with replacement.
  • The variance of the bootstrap replicates estimates the variance of the statistic.
  • Bootstrap can reduce variance compared to analytical methods when the analytical formula is unknown or biased.
  • Bootstrap does not reduce the variance of the data itself; it estimates the variance of a statistic.
  • Bootstrap may increase variance or be biased for small samples or dependent data.
  • Alternatives like the delta method or jackknife may be more appropriate in some cases.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.