0
$\begingroup$

given a linear regression model such as

$ y= \beta_0 + \beta_1X_1 + \beta_2X_2 + \beta_3D + \beta_4D*X_1 + \beta_5D*X_2 $

where $D$ is a dummy variable, what are the proper interpretations of the coefficients $\beta_1$ and $\beta_4$ ?

My understanding is that the partial base effect of a one unit change in $X_1$ will lead to a change in $y$ by $\beta_1$ units for both groups. (when $D=1 $ or $ D=0$)

and that there is an additional effect on $y$ of $\beta_4$ units only when $D=1$ given a one unit change in $X_1$

a classmate told me that $\beta_1$ only describes changes when $D=0$

$\endgroup$
1
$\begingroup$

You can get the marginal effect on $y$ from $X_1$ increasing by 1 unit by taking the partial derivative of $y$ with respect to $X_1$:

$$ \frac{\partial y}{\partial X_1}= \beta_1 + \beta_4 \cdot D$$

So the marginal effect depends on $D$, which is binary. For people who have $D=0$, the marginal effect is just $\beta_1=\beta_1 + \beta_4 \cdot 0$. For people with $D=1$, it is $\beta_1 + \beta_4 = \beta_1 + \beta_4 \cdot 1$. The effect of $X_1$ is modified by $D$ in this model.

If you are uncomfortable with calculus, you can also calculate a finite difference between the prediction if $X_1$ set to $k+1$ and the prediction with $X_1$ set to $k$.

| cite | improve this answer | |
$\endgroup$

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.