Couldn't you just have two different regression nodes, one for DV1 and one for DV2? I'm pretty sure that generates the same results as any PROC REG results you could get out of Base SAS + SAS/STAT. Then you'd have to write your own custom evaluation node.
If you want a truly multivariate analysis, where the correlations between DV1 and DV2 are taken into account in the modeling, you could use a Partial Least Squares node in EM (I have never done this in EM, but I know the method is in there and I assume it will allow multiple DVs).
... View more
