interaction term

Need to model interactions as interactions, or you get simpson's paradox

How do they show up? When we create our model, we might have some intuition that one variable gives context to another.

What should we do to explore the data if we have two continuous variables? Compute the conditional distributions. So let's say we have neuroticism and extroversion as variables and volunteering as a target. Then, we should hold neuroticism constant for a few selected values, and then look at extroversion vs. volunteering for each value.