Adaptation of Quasi-Least Squares to Estimate Correlations within a Nuclear Family
Roy T. Sabo, Narasinga Rao Chaganty · Communication in Statistics- Theory and Methods · 2009
In this article, we adapt the quasi-least squares, a robust procedure for estimating correlation between continuous variables, to the analysis of familial clustered data. The estimators derived from this procedure are compared with traditional maximum likelihood and moment estimators. The emphasis is on estimation of the correlations within a nuclear family. The quasi-least squares estimators are found to be nearly as efficient as the maximum likelihood estimators in the large sample case when the data are Gaussian. When the sample size is small, simulations show that the quasi-least squares estimators are more robust than the alternatives with respect to estimated infeasibility probabilities and mean square error. We conclude that the quasi-least squares procedure is preferable to the maximum likelihood and moment procedures in the analysis of correlations within a nuclear family.