0
$\begingroup$

I am reading Simple Regression Model from this book, Section 6.5 (page 267 in downloaded pdf, 276 if viewed online).

The author starts with below equation for a simple linear regression model,

$$ Y_i = \alpha_1 + \beta x_i + \varepsilon_i $$

And then after few lines, he lets for conveience that, $\alpha_1 = \alpha - \beta\overline{x}$ so that,

$$ Y_i = \alpha + \beta(x_i - \overline{x}) + \varepsilon_i $$

where $\overline{x} = \dfrac{1}{n}\sum\limits_{i=1}^nx_i$

My questions:
1. It is not convincing to bring in $\overline{x}$ just for convenience sake in the equation. Can any one please explain the logic behind bring that in the equation?
2. After above equation, the author says, $Y_i$ is equal to a nonrandom quantity, $\alpha + \beta(x_i - \overline{x})$, plus a mean zero normal random variable $\varepsilon_i$. Does that mean, $\alpha + \beta(x_i - \overline{x})$ has no randomness involved in that?

Kindly help.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$
  1. $\alpha_1$s in two equations are different. Let $\alpha_2$ be the $\alpha$ in the second equation, then $$ \alpha_1 = \alpha_2 + \beta \bar x$$

At the time that the computer was not popular or had no computer, the line was fit by using calculators. Bringing in $\bar x$ is really simplified the computation.

  1. From the first equation, $\epsilon$ is the only random component. So source of randomness of $Y$ is $\epsilon$, the other parts $\alpha + \beta x$ are known or unknown constant.
Improve this answer
$\endgroup$
3
  • $\begingroup$ I just corrected $\alpha_1$ to $\alpha$ in 2nd equation. Still the reason is not convincing that it simplified the computation. Can you kindly elaborate further? How could $\overline{x}$ suddenly enter the equation without an associated mathematical logic. $\endgroup$
    – Parthiban Rajendran
    Commented Oct 20, 2018 at 18:32
  • $\begingroup$ Let $z_i=x_i-\bar x$, then (1) $\sum z_i = 0$ vs calculating $\sum x_i$, (2) $\sum z_i^2$ is easier easier than $\sum x_i^2$, and (3) $\sum z_iY_i$ is easier easier than $\sum x_iY_i$. introducing $\bar x$ into equation does not change anything in equation, similar to $ + a - a$ , which we used to proof something in math. $\endgroup$
    – user158565
    Commented Oct 20, 2018 at 18:43
  • $\begingroup$ $Y_i = \alpha_1 + \beta x_i + \varepsilon_i$ ==> $Y_i = \alpha_1 + \beta x_i + \varepsilon_i - \beta \bar x + \beta \bar x $ ==> $Y_i = (\alpha_1 +\beta \bar x) + \beta (x_i - \bar x) + \varepsilon_i$ ==> $Y_i = \alpha + \beta (x_i - \bar x) + \varepsilon_i$ $\endgroup$
    – user158565
    Commented Oct 20, 2018 at 18:56

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.