Cross validation residuals were extended to generalised least squares models, showing the relationship with Cook's distance. For both ordinary and generalised least squares, there is no need to refit the model to smaller datasets for cross validation.