Suppose we construct an ensemble of k models by averaging their outputs. each of these models has error terms with a variance of v. assuming the worst-case scenario where the correlation between the errors is 1, what is the expected squared error of the ensembled model?