On Efficiency of Constrained Longitudinal Data Analysis Versus Longitudinal Analysis of Covariance