Robert Burrows <[EMAIL PROTECTED]> wrote in message
news:<a7vdpn$fia$[EMAIL PROTECTED]>...
> I have seen probability plots prepared by some statistics packages (i.e.
> Design-Expert and JMP) that have curved lines above and below the line fit
> to the data. What are these curved lines called and how are they
> calculated?
>
> TIA,
Sounds like prediction intervals (I assume the 'probability plots'
you refer to aren't Q-Q plots). These are basically confidence
intervals at each point of prediction. They're
calculated based on the distribution for the model error E
(in y=b_0+b_1*x+E) and the SEs for the Betas; i.e. CI for predicted value
y'=hat(b_0)+hat(b_1)*x',
at point x'.
I'm sure any decent regression book discusses this (Weisberg does).
Maybe SAS' online documentation has details.
HTH.
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
