Dear Anderson,
I appreciate your reply. Gender is a nuisance variable in this study which affects results significantly compared to a random sample with equal numbers of male and female in controls and patients. With this balanced data set, group comparison results are fairly robust regardless co-variances correction as expected except for age. So, I decide to apply two EVs in the two group comparison model correcting for age and gender in original data set. I wonder if it's good enough? BUT, Randomise is a non-parametric analyses, why it's sensitive to unbalanced data? Thank you in advance!
bw,
Joyce
|