STATISTICAL-ANALYSIS OF YIELD TRIALS WITH MATMODEL

Yield trials guide agronomic recommendations and breeding selections, but often are limited by inaccuracy, missing data, and the difficulty of understanding complex genotype-environment interactions. Recent studies have shown that a statistical model rather new to agriculturalists, the Additive Main Effects and Multiplicative Interaction (AMMI) model, can reduce these limitations. This paper introduces MATMODEL, a convenient program for the required calculations that runs on IBM compatible personal computers with MS-DOS (PC-DOS) 2.1 or higher and on the Apple Macintosh. MATMODEL enables one to (i) increase the accuracy of yield estimates, (ii) improve selections, (iii) impute missing data, (iv) model and understand the genotypes, environments, and interaction, particularly with a biplot graph, and (v) design flexible and efficient experiments. MATMODEL routinely provides yield estimates as accurate as raw treatment means based upon two to five times as many replications, so MATMODEL offers a remarkably cost-effective option for gaining accuracy. This option is particularly valuable in light of recent trends towards testing genotypes in more environments with fewer replications.