We introduce a fast stepwise regression method, called the orthogonal greedy algorithm (OGA), that selects input variables to enter a p-dimensional linear regression model (with p ≫ n, the sample size ...