David Morse , I am looking for some solid information about the interpretation of interactions in these non-parametric (aligned-rank transform) analyses. It is already difficult to understand what stochastic (in-)equivalence means in simple (one-way) designs, but I am clueless about the meaning when it comes to interactions between different factors. Do you have, by chance :), any literature you could point me to?
Do you have a problem/example of what you want to do at hand. I am not sure the protocol exist in R. But certain GLM in SPSS can help incorporate a covariate and give a robust output for data slightly failing parametric assumptions.
Hari Srinivasan variables cannot be parametric or nonparametric.
Nonparametric tests have tricky interpretations (as Jochen Wilhelm mentions) and most times have more assumptions than parametric tests, even in the univariate case.
You do not share much about your problem, but instead of plugging values into a package, wouldn't it make more sense to model your variable with a more appropriate tool such as a GLM, quantile regression or whatnot?