If you're asking whether between-subjects variance is greater or less than within-Ss variance, I'm puzzled that you need to ask: nearly always the between-SS variance is larger (often by an order of magnitude, so it doesn't take a lot of subtlety to notice it).
Ordinarily a simple repeated-measures analysis of variance (ANOVA) would explicitly display separate mean squares as estimates of these variances. Is there a reason why you haven't performed one? If the reason is that you don't (yet?) have different treatments to compare, so that your complete data set has 120 values (40 rows by 3 columns) and the putative "treatment" effect has 0 d.f. so your software balks at it, think of the design instead as a randomized-blocks (RB) design with Ss as blocks and your three repeated measures as treatments. A standard RB analysis will produce three mean squares: a 2-df "treatment" effect (for mean differences between the repetitions), a 39-d.f. "block" effect (the mean square for which is your between-SS variance estimate), and a 78-df effect probably labelled "Error" (which mean square is your within-Ss variance estimate). On Thu, 20 Nov 2003, Egil Ruefli wrote: > for each subject of my very small sample (n=40) I've got three repeated > measures (1 variable, 3 x 1 values). Thus there is a between-subjects > variance and a certain within-subjects variance. How can I find out > which of the two is higher? > > Thanks, Egil ----------------------------------------------------------------------- Donald F. Burrill [EMAIL PROTECTED] 56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816 . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
