2
$\begingroup$

Consider this model:

$$y = \beta_0 +\beta_1x_1 + \beta_2x_2 + \beta_3x_1x_2 + \varepsilon$$

Somebody told me today that the coefficient for the main effect of $x_1$ (i.e., $\beta_1$) will be 'marginal to the interaction'.

Could somebody explain what 'marginal to the interaction' means?

$\endgroup$
2
$\begingroup$

I wouldn't phrase it that way. The common meaning of "marginal" in statistics (and the way I use it) is roughly 'ignoring'. In other words, the marginal effect of $x_1$ is the effect of a 1-unit change in $x_1$ on $y$ ignoring all other possible variables. That is not what $\beta_1$ represents in that model. (Note that they may be using the terminology differently than I do, however; that happens.) To understand the idea of marginal (in the sense of ignoring) in regression better, it may help to read my answer here.

Instead, I would say the main effect of $x_1$ (i.e., $\beta_1$) in your model is the effect of a 1-unit change in $x_1$ on $y$ when $x_2=0$. To get a clearer sense of this, it may help to read my answer here (that answer in particular explains in detail what $\beta_1$ means when an interaction is present).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.