Interaction Effect

An interaction effect captures the idea that the effect of one variable depends on the level of another. You add it by multiplying two predictors together (x1 × x2) and giving that product its own coefficient, which then bends the model away from assuming each variable acts independently.

A classic example: a medicine might lower blood pressure a lot in older patients but barely in younger ones — that is an interaction between drug and age, and without the interaction term a plain model would force a single average effect on everyone. The cost is interpretation: once an interaction is present, you can no longer read a variable's main coefficient as its full effect, because the effect now changes depending on the other variable.