Assessment of research frameworks for on-farm experimentation through a simulation study of wheat yield in Japan
On-farm experiments can provide farmers with information on more efficient crop management in their own fields. Recent developments in precision agricultural technologies, such as yield monitoring and variable-rate application technology, allow farmers to implement on-farm experiments easily. Research frameworks including the experimental design, data collection and preprocessing steps, and the statistical analysis method strongly influences the precision of the experiment. Conventional statistical approaches (e.g., ordinary least squares regression) may not be appropriate for on-farm experiments because they are not capable of accurately accounting for the underlying spatial variations in a particular response variable (e.g., yield data). We explored the effect of sensor types, data preprocessing, experimental designs, and statistical approaches on the type I error rates and estimation accuracy through a simulation study hypothesized to conduct on-farm experiments in 3 wheat fields in Japan. Isotropic and anisotropic spatial linear mixed models were established for comparison with ordinary least squares regression models. The repeated designs were not sufficient to reduce both the risk of a type I error and the estimation bias. A combination of a repeated design and an anisotropic model is required to improve the precision of the experiments. Although the anisotropic model sometimes did not contribute to reducing the bias efficiently, the results of the anisotropic model showed large standard errors, especially when the estimates had large biases. This finding highlights an advantage of anisotropic models since they enable experimenters to consider the reliability of the estimates when needed.
READ FULL TEXT