On 8 Jul 2001 13:30:13 -0700, [EMAIL PROTECTED] (Juan Costa)
wrote:
> I have the same values obtained with two different operators... How I
> can compute the inter and intraoperator error in those set of
> values?...
What you are asking for is not entirely clear. If error is "inter"
or between two operators, then it is not "intra" or within.
Paired t-test shows the *mean* difference between operators
on a set of measures. We usually don't worry about a
difference in *variability* until someone draws our attention
to it (because of some special risk, or other reason for suspicion).
The Pearson correlation shows their agreement; the residual
error (the average of squared differences; or 1-r^2) shows
the other main aspect of difference.
I presume: that you need repeated measures by one operator
in order to get within-operator error. If that is all you have, you
can use the same procedures.
More complicated: If you have a larger design with
different sorts of replication, you can build a large ANOVA
model and look at the contributions to variance. The statistics
are computed with least difficulty or ambiguity if the design
has equal-Ns (balanced factors) throughout.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
