Rajarshi Guha wrote:
> Hi,
>   I have question regarding the interpretation of the standard error of
>   prediction in linear regression. As I understand this value is the
>   variance of the error term for the predicted value in question.
> 
> Is it possible to interpret this value as some sort of a reliability of
> the prediction? 

Reliability is not a word statistician would use in this particular 
instance. We would probably say it is a measure of the precision of 
the prediction.

> That is, a low value of SEP would indicate that the
> predicted value is more reliable (even though the residual might be high)
> compared to the situation with a high value of SEP?

A low SEP is more precise. A high SEP is less precise.

> If the above is not possible could anybody explain in what way the
> residual of a prediction is related to the standard error of the
> prediction?

Residual estimates the error at a single data point. SEP estimates 
the precision of the fitted line. These two are not the same.

Any particular measurement is made up of a true value plus some 
noise. Noise may be high for one measurement and low for another and 
almost zero for a third. However, when you fit a line, you are 
smoothing out all of the noise, to try to uncover structure. How 
well you uncover the structure is reflected in the SEP of a prediction.

-- 
Paige Miller
Eastman Kodak Company
paige dot miller at kodak dot com
http://www.kodak.com

"It's nothing until I call it!" -- Bill Klem, NL Umpire
"When you get the choice to sit it out or dance, I hope you dance" 
-- Lee Ann Womack

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to