Dear Ph.D. McLaren,
Thanks for your detailed explanation. I also met a problem when using the flexible factorial design. My data include two groups of subjects, similar to your example. What confused me is that such design could work in model specification and estimation steps, however, when I was asked to set the contrasts, I found that the contrast between conditions within the same group (e.g. G1C1 - G1C2, G:group, C:condition) and contrast of interaction between group and condition (e.g. (G1C1 - G1C2) - (G2C1 - G2C2)) could be put into the contrast manager, while contrasts for main effect of group (i.e. G1-G2), single regressors (e.g. G1C1) and comparisons between two groups of some condition (e.g. G1C1-G2C1) will be recognized as "invalid contrast". The contrasts I set were according to the "wighted contrast.pdf" you have shared and were in the same way as you have shown.
I found it strange because by using the full factorial design (factor group: independent variable, unequal variance; factor condition: dependent variable, equal variance) I could set all the above contrasts.
Have I missed something when I do the flexible factorial analysis?
Bests,
Dr. Sun Delin
Post-doctoral Fellow
Laboratory of Neuropsychology and Laboratory of Cognitive Affective Neuroscience, The University of Hong Kong
Room 622B, Knowles Building, The University of Hong Kong, Pokfulam Road, Hong Kong
|