On analysis of variance in the mixed model (Q762858)

In the paper the author combines some approaches to fixed and random effects models into a unified treatment suitable for study of the analysis of variance (ANOVA) in the mixed model \(Y=X\beta +e\), where e has the linear structure \(e=U_ 1\xi_ 1+...+U_ k\xi_ k\), where \(U_ i\) is a given matrix, \(U_ k=I\), \(\xi_ i\sim N(0,\sigma_ i^ 2I)\), \(\xi_ i\) and \(\xi_ j\) are independent, \(i\neq j.\) Conditions for an ANOVA are given. Implication for estimation of variance components from an ANOVA are also discussed. These results are largely an extension of work by \textit{F. A. Graybill} and \textit{R. A. Hultquist}, Ann. Math. Stat. 32, 261-269 (1961; Zbl 0109.375).