0
$\begingroup$

Suppose we have the regression model: $y_i$ = $\beta_0$ + $\beta_1x_i$ + $\epsilon_i$

where $y_i$ = ($Y_i$ - $\bar{Y}$) and $x_i$ = ($X_i$ - $\bar{X}$).

I need to determine if the regression line is guaranteed to pass through the origin.

This will be true iff $\beta_0$ = 0. We immediately see that $\beta_0$ = ($Y_i$ - $\bar{Y})$ - $\beta_1$($X_i$ - $\bar{X})$. Where $\beta_1$ is given by $\frac{COV(X, Y)}{VAR(X)}$. I don't believe this quantity is guaranteed to be 0, so would the answer be that we are unable to determine if the regression line passes through the origin?

-- Thanks

$\endgroup$
1
$\begingroup$

In general, if you are using the least squares coefficients and the regression model $y_i=\beta_0+\beta_1 x_i+\epsilon_i$, where $E(\epsilon_i)=0$ for all $i$, then the coefficient $\beta_0$ is given by $$ \hat{\beta}_0 = \overline{y}-\hat{\beta_1}\overline{x}.$$ If you also have $y_i = Y_i-\overline{Y}$ and $x_i=X_i-\overline{X}$ then you get $\overline{y} = \overline{Y}-\overline{Y}=0$ and similarly $\overline{x}=0$. All in all, $\hat{\beta}_0=0$, and as you say this implies that the regression line passes through the origin.

$\endgroup$

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.