On 10 May 2004 14:45:02 -0700, [EMAIL PROTECTED] (Paul R Swank) wrote: > The Pearson Correlation documents that the scores are in the same relative > order over time and the paired t-test determines if there is a shift in > means. A more direct approach is to use a generalizability analysis where > the relative error ICC is the reliability ignoring mean differences and the > absolute plus relative error ICC takes into account both. >
That works okay if you don't care to explore or can't fix the sources of mean differences. But it is pretty hard to compare to versions of ICC, to estimate whether there is evidence that the means have a problem. > Brennan, R.L. (1983). Elements of generalizability theory. Iowa City, IA: > ACT Publications. > > Cronbach, L.J., Gleser, G.C., Nanda, H., & Rajaratnam, N. (1972). The > dependability of behavioral measurement: Theory of generalizability for > scores and profiles. New York: John Wiley & Sons, Inc. (Cronbach, Gleser, > Nanda, & Rajaratnam, 1972). > [ snip, original post] -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
