among other things, it gives you a theoretical way of forming a
confidence interval for an individual's true score, formed by adding
and subtracting 1.96 times the SEM to the person's obtained score. 
this assumes you want a 95% interval, of course.  there are lots of
variations on this theme.  for example, there is such a thing as the
standard error of measurement of a mean, so that you can apply the
same logic to mean scores as to individual scores.

so lets say an individual gets a 60 on some instrument for which the
standard error of measurement is 5.  the 95% CI is roughly 50 to 70.
meaning what?  well, it's exactly the same interpretation as any other
confidence interval, except the concept is that of the sampling
distribution of a given individual's obtained scores formed by taking
independent random samples of items from a domain, administering each
to the individual and forming the sampling distribution with a mean
equal to the person's true score and a standard deviation equal to the
standard error of measurement.  if we were to take each individual
obtained score and form a 95% confidence interval for the true score,
95% of them would contain the true score for that individual.  we are
therefore 95% CONFIDENT that the particular one we have is one that
contains the individual's true score.


"esphinx" <[EMAIL PROTECTED]> wrote in message news:<[EMAIL PROTECTED]>...
> In what way does the standard error of measurement is related to test
> interpretation?
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================
  • SEM esphinx
    • JJ Diamond

Reply via email to