© 1972 by Biometrika Trust
An investigation of procedures for multiple-stage selection for a variate subject to errors of measurement
The University of Waterloo
Multiple-stage selection from an infinite population of varieties is studied. Measurements of the variate of interest are subject to error. The study makes use of an expression for the mean value of the desired variate in the selected population after any number of stages of selection. Weighted information from all completed stages is used in deciding whether to accept or reject a variety at a given stage. The optimality of the symmetrical operating conditions that have been suggested by other authors is studied for up to four stages of selection for a variety of selection schemes. The value of final acceptance of particularly promising varieties without waiting for the second stage of a 2-stage programme is also investigated. Finally, the effect of the presence of interaction on the optimum selection scheme is studied.
Key Words: Multiple-stage selection Errors of measurement in selection Multivariate normal integral Screening