Optimal covariance adjustment in growth curve models (Q1566701): Difference between revisions

We consider the selection of covariables for the covariance adjustment of parameter estimators in growth curve models. The procedure consists of obtaining the best subset of linear combinations of higher-order polynomials for minimizing the variance of the adjusted estimator of a particular linear combination of the lower-order polynomial coefficients. The coefficients of the required linear combinations are expressed in terms of the eigenvectors of appropriate matrices and the gain in precision due to covariance adjustment is measured by the magnitude of the corresponding eigenvalues. The results are illustrated numerically using data analyzed previously in the literature.