Wouter, I'm just a little confused by your use of "3x3x3=27" variables. Do you really mean that you have 3 factors, each at 3 levels? If so, each subject will have 3 responses...one for each factor... and each response can be at one of the 3 levels, is that correct? If so, are there 9 parameters to estimate in the variance-covariance matrix if an "unstructured" or multivariate type model is assumed? Warren May University of Mississippi Medical Center
"Wouter Duyck" <[EMAIL PROTECTED]> wrote in message news:<abvuhi$p9c$[EMAIL PROTECTED]>... > I need help with the following please: > > SPSS provides multivariate tests for repeated measures designs. This > approach is useful when the sphericity assumption is violated. However, > consider the following: a 3 x 3 x 3 design (all within-subject factors) with > 10 subjects > => 10 rows and 27 columns > > If i remember correctly, in a MANOVA, the following condition must be > satisfies: > N - M > K, with N = number of subjects; M = number of between subject vars > (not applicable); K = number op dependent vars. > > Hence, one shouldn't be able to carry out a MANOVA with 3 x 3 x 3 = 27 > dependent vars with just 10 subjects (rows). But: SPSS does it anyway. Can > anyone explain this? Are results given by SPSS not reliable? Or did I make a > mistake? > > > Tanx alot! > wouter . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
