Postby EViews Matt » Wed Jan 16, 2019 11:04 am
Hello,
The residuals will never be perfectly orthogonal. If you're still working with the synthetic w y model you've posted in other forums, you can even see that your randomly generated innovations e and u are not perfectly orthogonal (just execute "cov u e"). You can certainly apply an SVAR model, which will give you numerically orthogonal structural residuals (or very close). Just keep in mind that applying an SVAR model requires you to make additional assumptions about how your variables are related, so you're really "choosing" a model rather than "finding" a model.