John O. Rawlings

Applied Regression Analysis: A Research Tool.

Review of Simple Regression.- Introduction to Matrices.- Multiple Regression in Matrix Notation.- Analysis of Variance and Quadratic Forms.- Case Study: Five Independent Variables.- Geometry of Least Squares.- Model Development: Variable Selection.- Polynomial Regression.- Class Variables in Regression.- Problem Areas in Least Squares.- Regression Diagnostics.- Transformation of Variables.- Collinearity.- Case Study: Collinearity Problems.- Models Nonlinear in the Parameters.- Case Study: Response Curve Modeling.- Analysis of Unbalanced Data.- Mixed Effects Models.- Case Study: Analysis of Unbalanced Data.- Appendix Tables.- References.- Author Index.- Subject Index.

Order statistics for a special class of unequally correlated multinormal variates

The moments of the order statistics for independent normal variables are well known and used for many purposes including the determination of the moments of censored normal populations. Selection programs in plant and animal breeding routinely use the average of the expectations of the s top-ranking order statistics as the standardized selection differential in the prediction of genetic gain. In many cases, the assumptio~ of independence may not be satisfied as, for example, when the population under selection is composed of individuals with varying degrees of genetic relationship. The present study Was prompted by the need for some understanding of the effect of unequal correlations on the order statistics, ordering without regard to the genetic relationships. One abstraction of this problem is to assume that the sample of n individuals consists of m full-sib families with k