Chris Rohlfs

Picture this: You design an ordinary least squares (OLS) regression model and run it. Based on the results of that first model, you drop the statistically insignificant regressors and re-run. You repeat multiple times until you finally arrive at a model in which all the regressors (or the ones you care about) are statistically significant.

Some content is only viewable by ASA Members. Please login or become an ASA member to gain access.