Hi On Mon, 26 Jan 2004, Dennis wrote:
> Well, the discussion gets away. > > Perhaps I should not use 'contrasts' at all, because I took the word from > the table "within subjects contrasts" although I used polynomial. > > What about using effect: > > There was significant main effect in the within-subjects variable TASK, > F(1, 4) = 3.083, p = 0.018, which is has significant in the 4th order > EFFECT, F(1,1) = 5.997, p = 0.019. ? As Bruce noted, there was nothing wrong with your use of the term Contrasts. I was simply uncertain which kind of contrasts you had done (polynomial is only one of many different types), and Thom was concerned about the interpretability of a 4th order contrast if it was indeed polynomial. This is a concern because the pattern in the data for a 4th order contrast would have to be similar to +1 -4 +6 -4 +1 or its inverse. What are the odds you "expected" this pattern? Bruce also noted a possible problem with the df for the contrast being 1 in the denominator (1 is correct for the numerator). But in fact there also appears to be an error in what you reported above for the omnibus effect, F(1, 4) = 3.083. If the df = 1 for the numerator, then there are no effects to partition into polynomial or whatever contrasts. This was also what you reported in your initial message. Shouldn't the df for the numerator be 4 and something else for the denominator? To get a p = .018 for an F = 3.083, you would need something like df num = 4, df den = 124. Best wishes Jim ============================================================================ James M. Clark (204) 786-9757 Department of Psychology (204) 774-4134 Fax University of Winnipeg 4L05D Winnipeg, Manitoba R3B 2E9 [EMAIL PROTECTED] CANADA http://www.uwinnipeg.ca/~clark ============================================================================ . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
