Let's say fitted regression line is $y_i=b_1+b_2x_i$.That is, $b_1$ and $b_2$ are estimators.

The text book says that if $e_i$ (residual) is normally distributed, then $b_1$ and $b_2$ are also normally distributed.

The evidence given in the text book is that if the normality of $e_i$ implies the normality of $y_i$. Then, since $b_2=\sum w_iy_i$, $b_2$ is the sum of normal distribution, which is normal distribution.

But, there is no explanation with regard to $b_1$. The evidence given is $b_1=\overline y - \overline x b_2$. Can we prove that $b_1$ is also normally distributed from this?? or is there another way to show it?

Thank you in advance.


1 Answer 1


You can use the same reasoning as for $b_2$:

  1. $\bar{y}$ is Normally distributed since it's a sum of Normal distributions
  2. $b_2$ is Normally distributed as you suggest
  3. $\bar{x}$ is a number (i.e. a constant, your observations), which just scales $b_2$

So $b_1$ is the difference between two Normal distributions, which as you know, is a Normal distribution.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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