Linear regression interview questions
14 problemsCoefficients, R squared, noise and correlation. Expect to reason about what happens to a fit when a variable is added, rescaled, measured with error or replaced by pure noise. Each problem has hints and a worked solution, free, no account.
| # | title | level |
|---|---|---|
| 2 | When adjusted R² stops falling | medium |
| 11 | Two signals, one R² | medium |
| 14 | R² from pure noise | medium |
| 18 | Every row twice | medium |
| 29 | A regressor measured with noise | easy |
| 54 | Soft thresholding | medium |
| 70 | A regression line in a corner | medium |
| 91 | A flat line through a curve | medium |
| 146 | Three signals that overlap | medium |
| 170 | Two regression slopes | medium |
| 172 | Beta is not correlation | easy |
| 191 | Positive correlation, negative coefficient | medium |
| 258 | The missing regressor | medium |
| 278 | Two R² values, one joint fit | medium |
Other topics: probability · random processes · order statistics · inference · linear algebra · games · puzzles · all problems