Statistical Methods in Seed Corn Product Selection

Our purpose is to examine statistical methods, particularly field plot technique, hypothesis testing, and inference suggesting, from a seed corn (Zea mays L.) company viewpoint. Company managers must look beyond traditional statistical methods to develop systems that identify superior new products. These systems should mimic the processes used by farmers in selecting new hybrids that provide superior performance in the broad range of environments where they will be grown. This paper shows the evolution of seed company strategy to identify superior hybrids in product development and testing systems. Test resource allocation has changed to fewer replicates per location, to more locations per year, to more years per hybrid comparison, and to more data integration across test types. Experimental design has changed from sophisticated to simple designs, from experiment summaries to head‐to‐head comparisons (two hybrids compared in all tests in which both were grown) then to head‐to‐group comparisons (series of head‐to‐head comparisons for a hybrid with appropriate elite hybrids within a maturity group), and from analyses of variance to t‐test analyses. Stability regressions have become more popular. As improved statistical methods have been used, more genetic progress per unit of time has resulted; hybrids have become more stress tolerant, thus more widely adapted.