Would it be possible to implement correlated errors in a multivariate model like in this example http://jslefche.github.io/piecewiseSEM/articles/piecewiseSEM.html#an-example-using-piecewisesem (see contents 2.4)?
Have you taken a look at https://cran.r-project.org/web/packages/brms/vignettes/brms_multivariate.html? There you will find details on how to specify residual correlations in brms.
Yes, I did. But residual correlations are modeled for all the response variables (y1, y2, y3 using the example above). I am wondering if it would it be possible to model it only for y1 and y2 (like in the example). Many thanks!
I see. Right now, brms either estimates all or no residual correlations between response variables mostly because everything in between comes with a lot of technicalities when specifying the Stan code.
2 Likes
OK, thank you for the prompt reply.