Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
The design and analysis of agronomic field experiments with large numbers of treatments in spatially heterogeneous environments
Abstract
In the presence of spatial heterogeneity in experimental fields, the traditional random blocking has proven to be inappropriate and the ANOVA method of analysis inaccurate. The assumption that observations are independent is frequently violated because substantial correlation among data is generally present. Attempts to account for spatial variability in agricultural fields have considered both the experimental designs and the statistical methods issues. In the particular case of plant breeding trials, where plenty of seed of standard, well known varieties and not enough material, of less known new varieties are evaluated, the problem is specific. Augmented experimental designs and the random field approach to the analysis of such data were suggested. A comparative study of experimental designs under spatial heterogeneity was performed. Optimality criteria, A, A$\sp+$, D and REML estimates of variance components, were modified to adjust for the correlated-data situation. Four experimental designs were selected according to their optimality values. Two layouts having 48 treatments and a check and two having 56 treatments and a check were used. The results suggested that the check plots should be as close as possible to the other treatments. Methods of analysis were also compared. Simulated data sets for three cases of treatment effects (one group, two groups and eight groups of treatment effects) with two spatial covariance structures (spherical and exponential covariance models) and three different ranges (short, medium and long), were generated from the selected designs. These data were analyzed using four different models. The random field approach leads to more accurate estimation of the fixed effects of the model and the contrasts. The adjustment proposed for the variance matrix of fixed effects works better under the spherical than under the exponential covariance structure. The correction for the standard errors of estimation was pronounced more significantly in all situations.
Subject Area
Biostatistics|Statistics|Agronomy|Geography
Recommended Citation
Hadarbach, Driss, "The design and analysis of agronomic field experiments with large numbers of treatments in spatially heterogeneous environments" (1996). ETD collection for University of Nebraska-Lincoln. AAI9703777.
https://digitalcommons.unl.edu/dissertations/AAI9703777