Assumptions on a multiple linear regression model and elastic net I am interested in using elastic net regression in place of an multiple linear regression. I know when you perform a multiple linear regression you should check the assumptions such making sure the model has constant variance, normality of residuals, etc. 
My question is - if I am to use elastic net - do I have to check these assumptions for this as well?
Any comments and help would be appreciated! Thank you! 
 A: Strictly speaking, normality of residuals (Ei ~ N(0, sigma^2) is a formal assumption of the (classical, unregularized) multivariate linear regression model.  See e.g. equation 7.7 (p. 229) of Neter, Wasserman, and Kutner, Applied Linear Statistical Models, 3rd Ed.  The same text says (p. 247), "Box plots and normal probability plots of the residuals are useful for examining whether the error terms are reasonably normally distributed" for the same multivariate linear model.  One can still use a model for prediction if a sample dataset violates some of the model's assumptions.  Presumably in such cases the modeling process has yielded at least a satisfactory out-of-sample error rate.  In the case of a regression model, I'd want to see the shape of the projections of the model's response surface into one or two independent variables closely matching the shape of the same projections for the generating process, across all individual and paired independent variables.  Substantial departures from linearity would give me pause.
