“…Linearbilinear models could be graphically represented through biplots (Gabriel, 1971;Yan and Tinker, 2006) or enhanced biplots (Gauch, 2006), where genotypes and environments are plotted in a single graph. Later, mixed models were introduced to allow more flexibility and to model different correlation structures among environments (Piepho, 1998;Burgueño et al, 2008;Kelly et al, 2009;Cullis et al, 2010), increasing statistical power (Wang and Schmidt, 2002). A number of correlation structures can be modeled with mixed models, from a single correlation among environments (i.e., compound symmetry models; Piepho, 2005), heterogeneous variances (Piepho, 1998;Malosetti et al, 2004;Burgueño et al, 2008;Cullis et al, 2010), up to a unique correlation for each pair of environments (i.e., unstructured variance-covariance; Burgueño et al, 2008;Malosetti et al, 2013).…”