R-Squared Is Invariant Under Scaling

Reported in interviews at Two Sigma

Problem statement

Consider a univariate linear regression with intercept: $y_i = \beta_0 + \beta_1 x_i + \epsilon_i$. Suppose every data point $(x_i, y_i)$ is multiplied by 2, so the new data is $(2x_i, 2y_i)$.

Prove that the $R^2$ of the regression is unchanged. Explicitly show how SSE (sum of squared errors) and SST (total sum of squares) each scale, and demonstrate that the scaling cancels in the ratio.

Show hint

Which pieces of the regression output would you need to recompute on the doubled data in order to evaluate $R^2$?

Relevant roles: Data Scientist, Applied Scientist and ML Engineer.

Related free problems