Author Information

Matt Strand
Jim Higgins

Abstract

In many designed experiments in agriculture and the life sciences, a researcher can anticipate the direction that responses will take when treatments are varied. For example, in a 2-way factorial, a researcher may know that increasing the levels of nitrogen and phosphorus will increase yields of a crop. Classical analysis of variance does not take into account a known ordering among population means. However, it can be shown that by restricting the estimates of means to have the same ordering as the anticipated ordering of population means, a reduction in mean-squared errors of estimators will likely occur, often by more than 50%. A procedure used to create such estimates is called isotonic regression.

In this article, a recently proposed method of isotonic regression for lattice-ordered means will be presented after first reviewing well-established methods. The newer method will be illustrated using data from an entomology experiment. In addition, standard errors of the estimators will be approximated using a bootstrap procedure.

Keywords

lattice order; parametric bootstrap procedure

Creative Commons License


This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Share

COinS
 
Apr 25th, 10:35 AM

USING ISOTONIC REGRESSION TO IMPROVE ESTIMATION IN FACTORIAL EXPERIMENTS WITH ORDERED FACTOR LEVELS

In many designed experiments in agriculture and the life sciences, a researcher can anticipate the direction that responses will take when treatments are varied. For example, in a 2-way factorial, a researcher may know that increasing the levels of nitrogen and phosphorus will increase yields of a crop. Classical analysis of variance does not take into account a known ordering among population means. However, it can be shown that by restricting the estimates of means to have the same ordering as the anticipated ordering of population means, a reduction in mean-squared errors of estimators will likely occur, often by more than 50%. A procedure used to create such estimates is called isotonic regression.

In this article, a recently proposed method of isotonic regression for lattice-ordered means will be presented after first reviewing well-established methods. The newer method will be illustrated using data from an entomology experiment. In addition, standard errors of the estimators will be approximated using a bootstrap procedure.