I am attempting to run a MANCOVA of memory assessment data. My IV has three levels (no memory impairment, moderate memory impairment, and severe memory impairment). The covariate is education. The DV is 7 memory assessment scores, with the scores being recorded as either percentiles or scaled scores.
The problem I am encountering is meeting the assumption of univariate and multivariate normality with my data. I have several outliers (mostly because I am comparing patients with varying degrees of memory impairments on measures assessing memory). The IV groups are rather small (22, 26, and 27).
I have been using the DeCarlo SPSS Macro to assess multivariate normality. When I run the macro without transformed data, all of the subjects in my no memory impairment group have negatively skewed data ranging from (-.5 to -1.7) and all of the subjects in my 2 memory impaired groups have mostly positively skewed data ranging from (.6 to 1.03) for the moderately impaired and (.7-2.7) for the severely impaired.
Further the Omnibus test of multivariate normality is significant for all IV levels, and the omnibus test of univariate normality is significant for all but one of the DVs in the severely impaired group.
Any suggestions of transformations I can use to ensure my data is meeting the normality assumption?
