In detecting Multivariate outliers, many researchers recommend using Mahalanobis Distance (MD) technique. In SPSS, such function is readily available by assigning the Dependent Variable, and Independent Variables via regression analysis tool (Regression > Linear > Save > tick Mahalanobis). An output of MD scores will appear (MAH_1). Another new variable that reveals the upper-tail probabilities of MD is created by computing 1-CDF.CHISQ(mah_1, X). X refers to the number of predictors. Any cases with values below 0.001 indicate multivariate outliers.

In my case, I am testing 3 continuous Independent Variables, and 5 continuous Moderator Variables. The specific moderation tool used is the Conditional Process Analysis SPSS-Macro Model 1 (Hayes, 2013). My question is, what is the number of predictors (X) that I should use in running the MD test as described above to discover Multivariate outliers? Other than IVs, should I take into account moderators in the number of predictors?   

IMHO, moderators are considered as predictors that change the effect of IV on DV. But at the same time, there is a clear distinction between the effects of IV and MV (interaction), and as such MV is not a main predictor of interest (in predicting DV).