What is a drawback of using forward selection compared to backward selection in model selection?

Forward selection is a sequential variable selection approach that starts with no predictors in the model and adds them one by one based on specific criteria, such as statistical significance. A significant drawback of this method is its potential to miss variables that, while individually may not appear significant, could be jointly important with other variables in the model.

In the case of forward selection, the process evaluates variables independently for inclusion. This means that if two variables have a combined effect on the response but do not show a strong individual signal when considered separately, forward selection might overlook their importance. Conversely, backward selection starts with all potential variables and removes the least significant ones, allowing for a more comprehensive model that can capture these interactions and joint effects.

The other options do not accurately capture the limitations of forward selection. For example, it does not guarantee finding the best possible model since it is limited by the order in which variables are added. Additionally, forward selection does not inherently consider outliers more effectively than backward selection, nor is it restricted to categorical predictors since it can be applied to both categorical and continuous predictors.