Transition From ANOVA to AMMI Analysis to Improve Accuracy and Reporting of NTEP Data.

Data from the National Turfgrass Evaluation Program (NTEP) is used for making planting recommendations at the cultivar level based on field trials conducted in the USA and Canada. The NTEP Policy Committee voted to incorporate the additive main effect and multiplicative interaction (AMMI) model into data analysis of all new NTEP trials. AMMI satisfies a fundamental objective established by the NTEP Policy Committee, which is to enhance the quality and scientific merit of data analysis and reporting. The decision to transition from the standard analysis of variance (ANOVA) to AMMI is based on NTEP sponsored research including 2 years of modeling and 5 years of field validation comparing the accuracy of AMMI adjusted means with ANOVA cell means (means averaged over replicates). The use of AMMI by NTEP is a major departure from the competing ANOVA procedures. ANOVA presumes the cell means model is the most accurate in predicting turfgrass quality performance, which is the basis of cultivar recommendations. Unlike ANOVA, AMMI analysis requires two steps, (i) model validation to identify the optimal model and (ii), data fitting to compute AMMI adjusted means. AMMI achieves two important goals as it relates to NTEP objectives, (i) AMMI increases accuracy and simplifies recommendations by reducing the number of winners and (ii), AMMI assisted Mega-Environment Analysis (MGA) allows the ordering of NTEP test locations into homogenous groups according to interaction patterns. AMMI-MGA permits the same cultivars to be recommended to all locations of the same mega-environment. Further, means averaged over locations from the same mega-environment exhibit good predictive value for all locations, further simplifying cultivar recommendations. This is a major departure in the reporting of NTEP data from previous arbitrary groupings according to climatic, geographic or cultural factors and from previous policies that assume averaging over locations is unreliable for developing planting recommendations.