Two features in a dataset are perfectly linearly dependent (x₂ = 3x₁). What effect does this have on the rank of the design matrix?