Evaluation of the Mixed Linear Model with Orthogonalized and studentized Residuals

摘要

Residuals have been widely used to compare the fit of a model to the observations and to evaluate model assumptions. This paper develops orthogonalized and studentized residuals that adjust for the fit of the covariance structure in the mixed linear model. The resulting residuals are suitable for checking the model assumptions associated with the marginal, conditional, or random effect distributions. These residuals are also compared to Cholesky residuals and variance studentized residuals. An example is presented involving an experimental design with fixed and random effects.