Bruce - The crux of whether analysis of covariance is appropriate is not if you get group differences on activity when activity is the DV. That is no problem for the analysis of covariance you propose. What would be problematic would be if there was an interaction between acivity and group when predicting balance. If there is such an interaction, then the traditional analysis of covariance model represents a misspecified model and is not appropriate. In that case, you would want to formally model the interaction and then characterize it using procedures analagous to what Bruce suggests.
Jim Bruce Weaver <[EMAIL PROTECTED]> wrote in message news:<[EMAIL PROTECTED]>... > On Wed, 26 Feb 2003, DCD wrote: > > > I have collected some data using older adults have found a significant > > difference (p < .05) between the three groups in terms of their activity > > level. Subsequent to this analysis I am contrasting the three groups on a > > balance measure and deciding whether or not to use activity level as a > > covariate in the ANOVA calculations. > > > > The question I have is whether or not to use activity level as a covariate > > or not, based on the fact that the main effect for activity level violated > > the assumptions of the homogeneity of regression slopes when compared with > > one of the DV used in the study. I do have a small sample size and > > consequently did not use the covariate analysis but wanted to see if I was > > justified based on the violation listed above. > > > > Any help would be appreciated. > > > > Thanks, > > > > Clark Dickin > > > As I see it, there is nothing wrong with including the covariate per se. > The problem arises if/when you present adjusted means for only one value > of the covariate (i.e., predicted means for each group, usually at the > mean of the covariate), and pass them off as representative of the overall > variability among the groups. It would be more accurate to present a few > sets of adjusted means. For example, predicited means with the Covariate > equal to: 1) covariate mean; 2) covariate mean plus or minus 1 SD; 3) > covariate mean plus or minus 2 SD. Alternatively, you could get predicted > means at the quartiles. This kind of table or figure would show that the > variation among your groups depends on the value of the covariate. > > Cheers, . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
