Adjust YouTube Ad Scores Using Mixed-Effects Linear Regression
Overview
This question evaluates understanding of mixed-effects (hierarchical) linear regression and the ability to adjust for rater bias, testing competencies in statistical modeling, fixed-versus-random effects distinctions, and handling hierarchical rating data.
Google
Jul 12, 2025, 6:59 PM
Data Scientist
Technical Screen
Machine Learning
7
0
Scenario
- 100 reviewers each rate the same 100 YouTube ads on a 1–10 scale.
- Ratings may be systematically higher or lower for some reviewers (leniency/severity bias).
Goal
Produce unbiased ad scores (comparable on a common scale) by removing reviewer bias.
Task
Propose a modeling framework to adjust for reviewer bias and justify why a mixed-effects linear regression is appropriate.
Assumptions
- Each rating is on an interval-like scale (1–10). Treating it as approximately continuous is acceptable; see alternatives for ordinal modeling.
- We want ad-specific scores (so ads are fixed effects) and to treat reviewers as a sample from a larger reviewer population (random effects).
Solution
Login required
Show