Virtual StatsPD@Waite meeting

Every month, the professional development meetings of statisticians and data scientists at Waite, known as StatsPD@Waite, bring together specialists in various aspects of data sciences in agriculture from Waite, Roseworthy and Adelaide.

Please join us for the next Virtual StatsPD@Waite seminar where Michael Mumford from Agri-Food and Data Science, Queensland Department of Agriculture and Fisheries will present on incorporating environmental covariates in linear mixed models to account for genotype x environment x management interactions.

Please note that the StatsPD@Waite meetings are recorded. If you have a question to the speaker but would rather not be recorded, please send me your question via chat during the meeting and it will be asked on your behalf. 

Please email Beata Sznajder for details of the Zoom meeting.

A one-stage predictive linear mixed model for genotype x environment x management practice (GxExM) interactions incorporating environmental covariates  

Michael Mumford - Agri-Food and Data Science, Queensland Department of Agriculture and Fisheries

In field crops research, genotype by environment (G×E) interactions are ubiquitous. The development of statistical methods to model the genotype by environment interaction using environmental covariates (ECs) has mostly occurred within the context of crop breeding. In agronomic research, measuring the impact of management practice (M) on genotype performance is also a key objective, giving rise to the genotype by environment by management practice (G×E×M) interaction.

A one-stage analysis approach, implemented in a linear mixed model framework, is presented, including ECs to untangle the G×E×M interaction arising in a series of sorghum agronomy field experiments. The linear mixed model framework allows adjustments for design effects and spatial field trend, along with the estimation of heterogeneous residual variance across environments. Nineteen ECs representing known eco-physiological drivers of crop growth, development and stress were derived from weather records and crop observation within each environment. Covariate data were obtained at different strata, with some covariates measured at the G×E level, and others at the environment level. Covariates were incorporated into the model via subset selection, to identify the most important predictors, avoid multi-collinearity, and ensure a parsimonious model.

This study is the first step in developing one-stage statistical models that can assist in determining the key drivers of the G×E×M interaction, allowing for the development of more robust agronomic recommendations