This is simply the linear model we started with, but replacing the generic k subscript for
genotypes with 1, averaging over all other terms in the model, and dropping the term for residual
effects since they do not contribute to a marginal mean prediction. Realizing that the lm()
function sets the coefficients for