Algorithm 1: Subset Selection

1: Input: Training Data: ${\left(x_i^1, x_i^2, \dots , x_i^T;y_i\right)}_{i=1}^I$

2: For each $j=1, 2, \dots , J:$

3: (a) Fit linear regression model for all combinations $\left(\begin{array}{c}T\\ j\end{array}\right)$ of predictors.

4: (b) Set the best model BMj as the one with the highest coefficient of determination R2

5: Select overall best models (BMs) among BM1,BM2,…,BMJ as the one(s) that have adjusted R2 ≥ 0.99 and in which all predictors have a statistically significant relationship (p value ≤ 0.01) with the response