Say I have a multilinear model $y = X\beta - \varepsilon_1$, let's call this A.
I also have another multilinear model $y = X_i\gamma - \varepsilon_2$, let's call this B. Where $X_i$ is a subset of $X$, meaning that all the variables in B are also found in A.
According to some readings, as we increase the number of predictor variables, the SSR will almost always decrease. Can we prove this mathematically by showing that that $SS_{res}(A) \le SS_{res}(B)$.
