In a housing price model, floor area, room count and built-up area are almost perfectly correlated, and the coefficient estimates swing wildly between retrains. Which remedy addresses this redundancy?
Show answer and explanation
Answer: A
Near-duplicate predictors make coefficient estimates unstable. PCA rotates them into uncorrelated components, and removing one column from each highly correlated pair removes the redundancy directly. Duplicating columns worsens the instability, a larger step size does not change the underlying overlap, and an arbitrary choice may discard the more informative column.
Why the other options are wrong
- B. Duplicating a predictor gives the regression the same information twice, which makes the coefficients even harder to separate and more volatile.
- C. Optimizer settings control convergence speed, not whether the predictors overlap, so the swings in the coefficients would persist with a different step size.
- D. Choosing a column without checking can keep the weaker of two near-duplicates and lose signal, so it is a careless fix rather than a principled one.