MLA-C01 Practice Question: During data preparation for a regression model, a…

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Apply PCA to reduce dimensionality to a single component

  Why it's wrong here

  PCA produces linear combinations of both features, destroying their individual interpretability and discarding the variance each contributes separately, which regression coefficients then cannot express. PCA suits unsupervised dimensionality reduction on many correlated predictors, not a targeted fix for exactly two correlated features.

• ✓

  Remove one of the two features

  Why this is correct

  A Pearson coefficient of 0.95 between two features signals strong linear redundancy, so their coefficients become unstable and interpretation unreliable. Removing one feature eliminates the collinearity directly, satisfying the stem's multicollinearity concern while retaining the remaining feature's predictive information.

• ✗

  Keep both features as they are because linear models are robust to multicollinearity

  Why it's wrong here

  Ordinary least squares regression assumes no perfect multicollinearity; at 0.95 the coefficient estimates become unstable with inflated variance, so retaining both features harms the model. Keeping correlated features is defensible for tree-based models, which split on individual features and tolerate correlation.

• ✗

  Standardize both features using StandardScaler

  Why it's wrong here

  StandardScaler only rescales each feature to zero mean and unit variance; it leaves the 0.95 correlation between them completely unchanged, so the multicollinearity persists. Standardisation is correct when features differ wildly in scale and a distance- or gradient-based algorithm needs comparable magnitudes.

Go deeper

Related to this question