Author Information

David B. Marx

Abstract

Soil heterogeneity is generally the major cause of variation in plot yield data and the difficulty of its interpretation. If a large degree of variability is present at a test site, some method of controlling it must be found. Controlling experimental variability can be achieved either by good experimental design or by analysis procedures which account for the spatial correlation. Classical designs are only moderately equipped to adjust for spatially correlated data. More complex designs including nearest neighbor designs, Williams designs, and certain restricted Latin square designs are developed for field experimentation when spatial correlation causes classical designs to be less desirable

.

The designs, both classical and nearest neighbor type designs, are analyzed using the classical statistical analysis approach and a strategy using general linear mixed models which takes into account that there is spatial correlation present. The results indicate that properly designed experiments may be analyzed either by the usual statistical techniques or more complex methods which adjust for spatial correlation. However, if no serious thought is used in constructing the design of the experiment then the usual analysis techniques are no longer valid

.

Keywords

spatial variability, geostatistics, optimal designs

Creative Commons License


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

Share

COinS
 
Apr 26th, 12:00 PM

DESIGNED EXPERIMENTS IN THE PRESENCE OF SPATIAL CORRELATION

Soil heterogeneity is generally the major cause of variation in plot yield data and the difficulty of its interpretation. If a large degree of variability is present at a test site, some method of controlling it must be found. Controlling experimental variability can be achieved either by good experimental design or by analysis procedures which account for the spatial correlation. Classical designs are only moderately equipped to adjust for spatially correlated data. More complex designs including nearest neighbor designs, Williams designs, and certain restricted Latin square designs are developed for field experimentation when spatial correlation causes classical designs to be less desirable

.

The designs, both classical and nearest neighbor type designs, are analyzed using the classical statistical analysis approach and a strategy using general linear mixed models which takes into account that there is spatial correlation present. The results indicate that properly designed experiments may be analyzed either by the usual statistical techniques or more complex methods which adjust for spatial correlation. However, if no serious thought is used in constructing the design of the experiment then the usual analysis techniques are no longer valid

.