Analysis of Agricultural Field Trials in the Presence of Outliers and Fertility Jumps
Abstract
The object of most agricultural field trials is to assess the effect of treatment on yields. This must be done in the presence of unknown fertility trends through the field. Many stochastic models for these fertility trends have been proposed and shown to be efficient competitors to classical blocking designs. In this paper we shall restrict attention to trials where plot fertilities are correlated in only one direction. In the case of a rectangular lattice of plots this is often assumed when the long, thin nature of the plots makes those with common long edges spatially close compared to those with common short edges. When using a stochastic model for field fertility in the analysis of a particular trial two questions must be addressed. The first is the choice of fertility model, and the second is the treatment of anomalies such as outliers and fertility jumps, i.e. sudden levels shifts in fertility. Even when several such possibilities are entertained, it is not uncommon to condition on a single model and a single set of anomalies, which are determined by model diagnostics.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 18, 1991
- Accession Number
- ADA242454
Entities
People
- Adrian Raftery
- Ross Taplin
Organizations
- University of Washington