On 20 Dec 2002 05:36:16 -0800, [EMAIL PROTECTED] (Alfred Barron) wrote: [snip: less useful answer than those already posted, given to a question that was not cited... ]
> Now I have a related question. > Are there multiple comparison tests for time series data ? If I follow the question: No. The cogency of "multiple comparisons" comes from having the shared variance estimate for all comparisons. Where there are savings in power (the intent, except for Sheffe's), it is gained from assuming the comparisons are fundamentally alike - same N, same variance, same hypothesis - and they are 'correlated comparisons' owing to the computation of their shared variance. > Do we basically test with a form of repeated measures, > with, say, the covariance modeled by an AR structure, > and then compare means ? There are so many designs included under the label of time series data, that I can't guess what you ask here. The way I read it, "Do we basically test..." is redundant with "... and then compare means." -- 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/ . =================================================================
