Kaplon and Swank in their messages cleared up the problem. However, thanks
for all the comments.
If you define the "Confidence Interval" (about the mean of a sample) as:
Confidence Interval = t(p,n-1) * sqroot[ SSODFM / (n * (n - 1) )]
Where n is the sample size
t(p,n-1) is the t distribution value for a given p and n
SSODFM = Sum of the squares of the deviations from the mean,
the differences between texts (including Hogg and Craig) disappear. The
issue of whether the variance is biased or unbiased does not enter in here.
Variance in the combined equation does not show up. The issue of whether the
variance has n or n-1 degrees of freedom does not enter in. As Kaplon says,
you can argue either way, but the interval remains the same.
DAHeiser
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
