Bayesian Spatial Statistical Applications in Corn Planting Density Trials.

Monsanto Company conducted a series of field trials in 2008 to quantify the effects of planting density, row spacing, hybrid, and environment on grain yield and other traits of interest. For these trials, Bayesian regression models with spatially correlated location effects were used to make inferences. Markov Chain Monte Carlo simulations obtained by Gibbs sampling were used to describe the posterior distributions of the parameters of the models and related functions. By using Bayesian methodology instead of the more familiar frequentist routines for analysis of covariance, the inferences can be simplified for presentation to a broad audience and extended to include unconventional quantities, such as the expected change in yield with a small increase in the current planting density and the most likely realistic planting density that would maximize yield. Data from commercial breeding trials and from a study involving planters with variable seeding rates and row spacings will be used to illustrate the procedures and findings.