Hi On Wed, 24 Sep 2003, Christopher D. Green wrote:
> Assuming they are at least ordinal, you could code them 1, 2, > 3 (or -56, 7, 1006, if you wanted to). Although true that coding is arbitrary with 2 groups, this does not generalize to 3 groups. Depends on the hypothesized underlying "distance" between groups 1 and 2 and groups 2 and 3. > You'll have a restriction of range problem that will tend to > supress the final value you get for the coefficient, and the > "pct. of variance accounted for" interpretation of its square > will be of dubious value. With two predictors as described in my other posting (linear and quadratic or some other orthogonal set such as x1=-2 1 1, x2=0 -1 +1), the overall R^2 will be equivalent to eta^2 and the individual r^2 for each predictor will be the percentage of variability attributed to linear and quadratic components (or contrast 1 and contrast 2 more generally). Latter will add up to overall R^2 (unequal ns may affect this). 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 ============================================================================ --- You are currently subscribed to tips as: [EMAIL PROTECTED] To unsubscribe send a blank email to [EMAIL PROTECTED]
