Restriction is particularly useful in the following scenarios:
When the confounder is strongly associated with both the exposure and the outcome. In early stages of research when simple and direct methods are preferred. When the restricted category is large enough to maintain statistical power.
