Estimation of parameters in the growth curve model with a linearly structured covariance matrix -a simulation study

Abstract

In a study made on estimation of parameters in the extended growth curve model with linearly structured covariance matrix [5] through simulation for some linear structures it was noted that the estimates of the covariance matrix may not be positive de nite for small sample sizes whereas it is always positive de nite for some other structures. In this master thesis the implementation of the algorithms proposed in [6] and [5] for some known linear structures on the covariance matrix was performed and simulation study for di erent small sample sizes were repeated many times. For these simulations the percentage of non positive de nite estimates were produced and the linear structures that always produce the positive de nite estimates were identi ed and classi ed from all those linear structures studied.