Iterative Nonorthogonal Analysis of Covariance

Abstract

This paper develops a method for handling a nonorthogonal analysis of covariance in an iterative manner using balanced analysis of variance residual and expectation operators. In essence, it extends previous work of the author for the nonorthogonal AOV problem to the nonorthogonal AOC problem. The iterative AOC method has the property of guaranteed convergence. In addition, under certain (convergent) conditions, successive approximations to the residual sum of squares for the AOC model are shown to be monotonically decreasing. This property is used to minimize iteration in hypothesis testing.