Sr. Searle, BUILT-IN RESTRICTIONS ON BEST LINEAR UNBIASED PREDICTORS (BLUP) OF RANDOM EFFECTS IN MIXED MODELS, The American statistician, 51(1), 1997, pp. 19-21
In the usual mixed model of analysis of variance we show that certain
sums of best linear unbiased predictors (BLUP) of random effects are z
ero. Those sums are similar to, but not exactly the same as, those of
the Sigma-restrictions sometimes used for fixed effects.