A house-price model fit on the Ames data gives: price = 45,000 + 112 × living_area_sqft + 9,800 × garage_cars − 650 × house_age. All features are in their natural units.
How should the coefficient on garage_cars be read?
Why did you pick that answer? Two or three sentences. The act of articulating it is what builds the judgment — not the click that follows.
A linear coefficient is a slope in the feature's own units: one more garage space, all else equal, moves the prediction by $9,800. It is a statement about the model's predictions, not a causal claim, and its size cannot be compared to a coefficient in different units (a square foot is a much smaller unit than a garage bay).
112 per square foot times a typical 1,500 sq ft is $168,000 of predicted price — living area moves the prediction far more. Compare coefficients only after standardizing the features.
The model describes how prices vary with garages in this data. Houses with big garages also tend to be newer and larger; the coefficient is an association, not the effect of building a garage.
Variance explained is R², a property of the whole model. A coefficient is a slope, not a share.