Looking at the documentation, I think I understand what “fitted” and “predict” mean for the residual computation method, but want to make sure.
Fitted are the differences between the actual data and the regression line. Predict are the differences between the actual data point and a “fake” data point drawn from the likelihood conditioned on the value of the covariates? So predict residuals have added likelihood noise in them?
- Operating System: Mac OS X
- brms Version: 2.9.0