Readers familiar with fixed-effects regression models may be concerned that the effects posited in this model are not estimable. Such a concern is well placed-the individual effects cannot be estimated. Furthermore, regression methods cannot deliver unambiguous estimates of the relative importance of classes of effects. However, the statistical problem is not to estimate the thousands of effects, but to estimate the six variances. Despite the nesting in the model, the variance components are estimable. Note that it is the assumption that the underlying effects are realizations of random processes that allows a measure of their relative 'importance'. Were they assumed to be 'fixed', one could test for statistical significance, but there would be no reliable way of assessing importance. It is only by estimating the variances of effects that relative importance can be assessed.