Individual analyses of variance were performed for each year
as described by Petersen (1985) and Scott and Milliken (1993)
with genotypes as fixed effects and blocks as random effects. The
inbred check PB130 was removed from the data analysis due to its
extremely high levels of fumonisins and a large standard error. The
two-way analysis of random year effects and fixed genotype effects
was performed with the least square mean estimates of treatments
obtained for each year (Federer et al., 2001). Mean comparisons
among genotypes were computed using Fisher’s protected LSD.