04-03-2017 02:59 AM
I am having an issue trying to find a way to code a nonparametric ANCOVA, and I am wondering if its even possible in SAS. I have 1 fixed effect and 1 covariate. The assumption of normality is met, however the assumption of homogeneity of errors is not met (p-value for fixed effect = 0.0432 using Levene's test). I have tried several transformations (log, ln, power, squareroot, arcsine squareroot, etc.) and have not been able to meet the assumptions. I was hoping to do a ranked analysis, however I have not been able to figure out how to code it. Any assistance would be greatly appreciated!