Which method of input selection determines if the total model improves on the baseline as variables are added?

The method of input selection that assesses whether adding variables improves the model's performance compared to a baseline is known as forward selection. In this approach, the model begins with no predictors and sequentially adds variables one at a time, evaluating the model's performance with each addition. The process continues as long as the new variable significantly improves the model, which is typically measured through criteria such as the AIC, BIC, or adjusted R-squared, or by comparisons to a baseline model without that variable.

This method is particularly advantageous when the number of potential predictors is large and helps identify which variables significantly contribute to explaining the variability in the response variable. The focus is not just on adding any variable, but rather on those that genuinely enhance the model compared to an existing baseline, ensuring efficient and effective model development.

In contrast, other methods such as backward selection start with all candidate variables and remove them based on certain criteria, while stepwise selection incorporates both adding and removing variables based on performance metrics. Sequential selection is a broader term that can sometimes refer to processes that go beyond simple addition of variables at each step. However, the specific combination of adding variables and directly comparing against a baseline aligns perfectly with the principles of forward selection.