This is the joint presentation by me and Suman at SAGI Symposium 2020.
Part 1 is given by Suman about OFE and GWR approach. The detail is at publication.
Title: On-farm Strip Trials: going beyond small plot experiments
Abstract: Accommodating spatial variation is common in analysing field trials, and has become a challenge in large on-farm experiments. Simple linear mixed models and Gaussian distribution are incapable of dealing with complex data sets and results in misspecification. This paper describes a Bayesian framework to analyse the model with spatially correlated random parameters for large on-farm experiments. With advanced model diagnostic tools, we found that the model with Gaussian assumption is misspecified. Therefore, we use Student distribution with which the model is augmented. The open-source R-code that implements our proposed method for analysing on-farm data is provided for further use. We also discuss the difference of the Bayesian approach and GWR, and compare the results from these two approaches.
