Consider a simple linear regression model $Y = X\beta + \epsilon$ $ \ $. Let $Y_O$ be the least square predictor of $Y$ at $X = x_o$ , based on $n$ observations $ (X_i ,Y_i) $and $\overline X$ = sample mean for $ X_i $

then the standard error of the predictor $Y_O$

a) increases as $ x_o $ moves away from $\overline X$

b) decreases as $ x_o $ moves closer to 0

This is what I tried :- according to me the answer must be option a) because the standard error of the predictor is given by

$s\sqrt {\frac 1n + \frac {(x_o - \overline x )^2}{\sum {(x_i - \overline x)^2}}}$

so clearly if $ x_o $ moves away from $\overline X$ standard error increases, but the source where the answers are given (without explanation) says option 'b' is correct.

I am not able to understand how that is. I think it's a misprint, but still I wanted to confirm if from any approach option b) is more relevant than a)


1 Answer 1


Your formula, reasoning and conclusion are right, so as you say, a is correct, b is (demonstrably) wrong.

Unfortunately there's not much to add; only if $\bar{x}$ were at $0$ would b be correct, but it would simply say the same thing as a in a different, less general way.


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.