In most breeding applications (and in all cases in this book) elements of y can be assumed as drawn from a normal distribution. Keep in mind that although this is a rather convenient assumption that facilitates the analysis of the data, it is not always tenable. We have mentioned in the previous section how we will consider both fixed and random effects in our models. We will now provide a short explanation on what we might consider fixed vs. random. The distinction between fixed and random applies to the unknown model components. A fixed effect is a known constant that will remain the same over conceptual repeated sampling, while a random effect is a random variable that arises from the subsampling and random selection of “treatment” levels.
Imagine a very simple fixed effect model similar to the following: