The problem of determining optimal designs for least squares estimation is considered in the common linear regression model with correlated observations. The approach is based on the determination of ‘nearly’ universally optimal designs, even in the case where the universally optimal design...