PROGRESS in plant and animal breeding is often

Transcription

1 Copyright Ó 5 by the Genetics Society of America DOI:.534/genetics Optimal Allocation in Designs for Assessing Heterosis From cdna Gene Expression Data Hans-Peter Piepho Institut für Pflanzenbau und Grünl, Universität Hohenheim, 7599 Stuttgart, Germany Manuscript received November, 4 Accepted for publication May 3, 5 ABSTRACT Heterosis is defined as the superiority of a hybrid cross over its two parents. Plant animals breeders have long been exploiting heterosis, but the causes of this phenomenon are as yet only partly understood. Recently, chip technology has opened up the opportunity to study heterosis at the gene expression level. This article considers the cdna chip technology, which allows assaying two genotypes simultaneously on the same chip. Heterosis involves the response of at least three genotypes (two parents their hybrid), so a chip or microarray constitutes an incomplete block, which raises a design problem specific to heterosis studies. The question to be answered is how genotype pairs should be allocated to chips. We address this design problem for two types of heterosis: midparent heterosis better-parent heterosis. The general picture emerging from our results is that most of the resources should be allocated to parent-hybrid pairs, while chips with parent-parent pairs or hybrid-reciprocal pairs should be used sparingly or not at all. PROGRESS in plant animal breeding is often made by exploiting nonadditive gene action. For example, when two maize inbred lines are crossed, the resulting hybrid is frequently found to be superior to the midparent value, i.e., the average of the two parent means (Falconer Mackay 996; Lynch Walsh 998). This phenomenon is commonly denoted as midparent heterosis of hybrid vigor. Historically, heterosis was first studied at the phenotypic level of agronomically relevant traits such as yield. Several theories have been put forward to explain heterosis (e.g.,stuber et al. 99), but a consensus has not yet emerged. The advent of chip technologies has now opened up the scope to study heterosis at the gene expression level (Ni et al. ; Kollipara et al. ; Guo et al. 3), thus increasing our understing of the underlying molecular basis of heterosis (Birchler et al. 3). This article is concerned with the optimal design of gene expression studies aiming at heterosis. The notion of heterosis may be associated with a linear model as follows. The expected phenotypic values of two parent genotypes A B their hybrid AB can be expressed as m A ¼ f t A ; m B ¼ f t B ; m AB ¼ f t AB ; ðþ ðþ ð3þ Address for correspondence: Institut für Pflanzenbau und Grünl, Universität Hohenheim, Fruwirthstrasse 3, 7599 Stuttgart, Germany. piepho@uni-hohenheim.de where f is a general effect t is the genotypic effect. Midparent heterosis may be defined by the linear contrast d AB ¼ m AB ÿ m A m B ¼ t AB ÿ t A t B : ð4þ Midparent heterosis occurs whenever d AB 6¼. Often, it matters which inbred line is the male parent. It is then important to also study the reciprocal cross, which we denote as BA. The linear model for this genotype is m BA ¼ f t BA : The reciprocal s midparent heterosis is ð5þ d BA ¼ t BA ÿ t A t B : ð6þ Heterosis of an agronomic trait is economically useful, when the hybrid outperforms both parents. This type of heterosis is also known as better-parent heterosis, it will occur for hybrid AB, whent AB. t A t AB. t B, assuming nonnegative coefficients that an increase in average phenotype is considered advantageous. Heterosis is thought to be associated with nonadditive gene action or dominance. In fact, dominance may be regarded as midparent heterosis at the gene level. Similarly, overdominance occurs when there is betterparent heterosis at the gene level. If an expression product for a specific gene can be measured for the inbred parents their hybrids, dominance can be estimated on the basis of (4) (6). Similarly, overdominance can be assessed on the basis of the contrasts t AB ÿ t A t AB ÿ t B at the expression level. This article is concerned with cdna chip technology, where each of a large number of genes is represented by Genetics 7: (September 5)

2 36 H.-P. Piepho a cdna spot on a glass slide. Expression profiles of two mrna samples representing two different genotypes are assayed on a slide in parallel. Genotypes are labeled by fluorescent dyes, resulting in a green signal for the one genotype a red signal for the other genotype. To account for dye effects, it is customary to swap dyes on about half of the chips assigned to the same genotype pair. Statistically, a microarray may be considered as an incomplete block accommodating only two treatments (genotypes) (Kerr Churchill ; Kerr 3). The design problem is how to allocate different genotype pairs to chips. Most of the current literature on experimental designs for identifying differentially expressed genes deals with the case where two or more treatments of equal interest are to be compared. Efficient designs in this context are the reference design, the loop design, balanced block designs (Dobbin Simon ; Kerr 3; Dobbin et al. 3a,b; Simon et al. 3). The objective of heterosis studies differs from those commonly considered in that the treatment contrast of interest involves three treatments, so efficiency regarding all pairwise comparisons is irrelevant. Also, most of the theory of optimal designs revolves around criteria such as A-optimality or E-optimality ( John Williams 995; Yang et al. ), which strive for optimality relative to a broad class of contrasts. In the case of heterosis, such approaches are not optimal, because the class of contrasts of interest is much more limited. Clearly there is only one type of contrast. While other designs such as the loop design may provide good heterosis estimates (Gibson et al. 4), they are not usually optimal (Keller et al. 5). By analogy, a balanced block design optimal with respect to all pairwise comparisons is not optimal regarding multiple comparison with a control. Generally, it will be more efficient to directly optimize the design with respect to the particular contrast(s) of interest ( John Williams, 995). This article is concerned with the problem of finding a design by which heterosis or dominance can be estimated with minimal stard error. Specifically, we search for the optimal allocation of a fixed number of chips among all possible genotype pairs. We first consider midparent heterosis then turn to better-parent heterosis. With both types of heterosis, we study the case of two hybrids as well as that of a single hybrid (no reciprocal tested). The derivations for different cases are organized as follows: first an appropriate linear model is formulated contrasts of interest are defined in terms of the parameters of that model. Optimality is then defined in terms of the variance of a contrast of interest. Minimization of this criterion leads to the optimal allocation. MIDPARENT HETEROSIS Hybrids reciprocal: We assume that analysis of normalized gene expression data is done in stard fashion on the basis of a linear model for log measurements. The model accounts for all relevant effects, including dye, chip, genotype (treatment). For details the reader is referred to Kerr Churchill (), Wolfinger et al. (), Keller et al. (5). It is assumed throughout that chip effects are taken as fixed, implying that interchip information is not recovered. This approach corresponds to the usual assumption made when deriving optimal incomplete block designs ( John Williams 995). Since there are only two genotypes per chip, all information on genotype contrasts is contained in pairwise differences of genotypic expression levels per chip. Clearly, the analysis of differences of log measurements is equivalent to analysis of actual log measurements, when chip effects are fixed. We express the model in terms of genotype differences, because this greatly simplifies our study of optimal allocation. In applications one will not usually analyze actual log intensities instead of differences. Let y ji denote the ith observed genotype difference for the jth genotype pair. Specifically, let y i ¼ ith observation (chip) on difference A ÿ B (i ¼,..., n ), y i ¼ ith observation (chip) on difference A ÿ AB (i ¼,..., n ), y 3i ¼ ith observation (chip) on difference A ÿ BA (i ¼,..., n 3 ), y 4i ¼ ith observation (chip) on difference B ÿ AB (i ¼,..., n 4 ), y 5i ¼ ith observation (chip) on difference B ÿ BA (i ¼,..., n 5 ), y 6i ¼ ith observation (chip) on difference AB ÿ BA (i ¼,..., n 6 ), where n j is the number of chips used for the jth genotype pair. The differences have the following expected values: Eðy i Þ¼t A ÿ t B Eðy i Þ¼t A ÿ t AB Eðy 3i Þ¼t A ÿ t BA Eðy 4i Þ¼t B ÿ t AB Eðy 5i Þ¼t B ÿ t BA Eðy 6i Þ¼t AB ÿ t BA : ð7þ The total sample size is given by n ¼ P 6 i¼ n i. For symmetry reasons, we require the same number n of observations for each parent-hybrid pair, i.e., n ¼ n 3 ¼ n 4 ¼ n 5 ¼ n. Thus, the optimal allocation is given by (n, n, n 6 ). To ensure identifiability, we set t BA ¼. To account for dye effects, one commonly swaps dyes for half the chips of a genotype pair. The dye swap can be accommodated by extending the linear model with dye effects dye-by-genotype interactions. To derive a design optimal with respect to contrasts among genotype main effects, it suffices to use model (7) require

3 Optimal Allocation in Designs for Assessing Heterosis 36 that the number of arrays for a particular genotype pair be allocated in equal parts to both possible dye swaps. Model (7) may be expressed as EðyÞ ¼X b; ð8þ where y ¼ ðy ;... ; y n ; y ;... ; y n ;... ; y 6 ;... ; y 6n6 Þ9, b ¼ ðt A ; t B ; t AB Þ, X is the appropriate design matrix with dummies ÿ. The heterosis contrast of BA can be written as d BA ¼ k9 dðbaþ b, where k dðbaþ ¼ ÿ A ¼ÿ The least-squares estimator is which has variance dˆba ¼ k9 dðbaþ ðx 9X Þ ÿ X 9y; varðdˆba Þ¼k9 dðbaþ ðx 9X Þ ÿ k dðbaþ s ¼ 4 9Ds ; where D ¼ I n n 6 n n n n n n n 6 n : ð9þ J ðþ ðþ ; ðþ I is a 3 identity matrix, J ¼ 9 with ¼ (, )9, s is the variance of a difference y ji (j ¼,..., 6). A derivation of Equation is given in the appendix. A design for a given sample size n involves an allocation (n, n, n 6 ) to the different genotype pairings. ÿ We now derive an allocation ÿ that minimizes var dˆba.it is first shown that var dˆba does not depend on n. Thus, we set n ¼. In the next step, we find the optimal value of n subject to the constraint n ¼ 4n n 6. It can be shown that 9D ¼ n n 6 n n 6 n ; ð3þ which is free of n. Thus, for any fixed values of n n 6, the variance of the heterosis contrast does not change with n. This proves that parent-parent chips (A-B pairs) do not add any information with regard to the heterosis contrast (6), so the optimal design must have n ¼. Setting n ¼ n 6 ¼ n ÿ 4n,we obtain o9d on 9D ¼ n ÿ n n ðn ÿ 3n Þ ð4þ ¼ÿ n ðn ÿ 3n Þ ÿ ðn ÿ n Þðn ÿ 6n Þ n ðn ÿ 3n Þ ¼ ; ð5þ yielding a quadratic equation in n with roots n ¼ r ffiffiffiffiffiffiffiffiffiffiffi 6!n 4 ÿ ¼ r ffiffiffiffiffi! 6 6 n: ð6þ Since n # n/4, the only feasible solution is n ¼ r ffiffiffiffiffi! ÿ n, 4 n: ð7þ Thus, for a given total sample size n, the quantity 9D, ÿ hence the variance of the heterosis estimator,, is minimized for the allocation var dˆba n ¼ r ffiffiffiffiffi! ÿ n :3n: ð8þ This same allocation also minimizes the variance of the other heterosis contrast (4), dˆab. The optimal allocation was derived by looking at a single gene, while in gene expression studies thouss of genes are studied simultaneously. It is perhaps useful to point out that generally the optimal allocation derived here is independent of the variance s, which may be gene specific. Thus, the optimal allocation applies to all genes simultaneously. Differences among genes in variance affect only the optimal total sample size needed to achieve a desired accuracy, which may be determined by stard procedures (Steel Torrie 98). Only one hybrid: When only one of the two possible hybrids is tested (hybrid AB, say), the model simplifies to EðyÞ ¼ Xb~; ð9þ with b ¼ðt A ; t B Þ9 the constraint t AB ¼. It can be shown that D ¼ð X9 XÞ ÿ ¼ I n J ; ðþ n n n where n is the number of A-B pairs n is the number of A-AB or of B-AB pairs. Noting that d AB ¼ k9 dðabþ b~ with k dðabþ ¼ÿð=Þ, it can be shown that varðdˆab Þ¼ n s ; ðþ i.e., the variance does not depend on n, the number of parent-parent (A-B) pairs. Obviously, the variance is minimized when n ¼ n/, where n is the total sample size. Thus, all microarrays should be allocated to parenthybrid pairs. Additive gene effects: In deriving an optimal allocation, we have focused on the accuracy in estimating d. It is sometimes of interest to also estimate the additive gene effect. The accuracy of such estimates in designs

4 36 H.-P. Piepho optimized for d is now considered for the two cases studied (Hybrids reciprocals as well as Only one hybrid). Design for hybrids reciprocals: By not allocating any chips to A-B pairs, we have no direct comparisons of the parents. It turns out, however, that the A-B comparison can be made with good accuracy. More specifically, it may be of interest to estimate the additive gene effect defined by where a ¼ t A ÿ t B B k a ¼ k a b; ÿ ðþ C A: ð3þ The additive gene effect is of interest when studying the mode of dominance. When d ¼ a, there is complete dominance, while dominance is only partial when d, a overdominance occurs when d. a (Kearsey Pooni 996). To study the mode of dominance it is desirable to estimate a with about the same accuracy as d. It turns out that with n ¼ we have varðâþ ¼k9 a Dk a ¼ s ; ð4þ 4n so that from () (3) varðdˆþ varðâþ ¼ n ÿ n. for n, n=4: ð5þ n ÿ 3n So generally, the additive genetic effect a will be estimated more accurately than both d AB d BA, when the design is optimized with respect to these two heterosis contrasts. Design for only one hybrid: The additive gene effect a BA ¼ k9 a b~ with k a ¼ð ; ÿ Þ9, is estimated with variance varðâþ ¼ ðn n Þ s : ð6þ Thus, when all microarrays are allocated to parenthybrid pairs (n ¼ ), the additive effect is estimated with the same accuracy as the dominance effect. BETTER-PARENT HETEROSIS Hybrids reciprocal: It is most convenient to consider the hybrid BA. Results for the other hybrid, AB, are analogous. Assessing better-parent heterosis of hybrid BA requires good estimates of the contrasts l BA(A) ¼ t BA ÿ t A l BA(B) ¼ t BA ÿ t B. The coefficient vector for the first of these contrasts equals k9 lðba;aþ ¼ ðÿ; ; Þ, the associated variance is varðtˆba ÿ tˆaþ ¼k9 lðba;aþ Dk lðba;aþ ¼ D ¼ 3n n n 6 n n 6 n n ; ð7þ 4n ðn n Þðn 6 n Þ where D is the first diagonal element of D in Equation. The variance D is seen to be symmetric in n n 6 ; i.e., the equation remains unaltered if n n 6 are exchanged. Therefore the optimal design should be such that n ¼ n 6. The common sample size is denoted as n, i.e., n ¼ n 6 ¼ n ; whence D ¼ 3n n 4n n 4n ðn n Þ : ð8þ After some algebra using n ¼ (n ÿ 4n )/ this becomes D ¼ n n 4n ðn ÿ n Þ : ð9þ The differential equation od =on ¼ yields a quadratic equation in n, which can be shown to have roots n ¼ ÿ r ffiffiffi! 6 n: ð3þ Obviously, the only feasible solution is n ¼ ÿ r ffiffiffi! n :7n: ð3þ Thus, % of the total sample size is to be used with each of four parent-hybrid pairs, leaving a little,% for the parent-parent pair A-B the hybrid-reciprocal pair AB-BA. Asn ¼ n 6 for the optimal design, % should therefore be allocated to each of these two pairings. As in the case of midparent heterosis, most of the resources (8%) should be used on the hybridparent pairs. Only one hybrid: The variance of the contrast l AB(A) ¼ t AB ÿ t A is varðtˆab ÿ tˆaþ ¼ D ¼ n n ðn n Þn ; ð3þ where D is the first diagonal element of D in Equation. Using n ¼ n ÿ n, this can be shown to equal D ¼ n ÿ n ðn ÿ 3n Þn : ð33þ Maximization again leads to a quadratic equation in n, which has roots r ffiffiffi! n ¼ 6 n: ð34þ 3 The only feasible solution is therefore given by n ¼ r ffiffiffi! ÿ n :465n: 3 ð35þ

5 Optimal Allocation in Designs for Assessing Heterosis 363 Thus, 84% of the sample size is allocated to the parenthybrid pairs, while only 6% of the chips are spent on the parent-parent pair. It is worth mentioning that the design problem here is equivalent to that of a multiple comparison with a control. For a completely romized design when two treatments are to be compared with a control, the optimal allocation is known to be m h =m p ¼ ffiffiffiffiffi :44, q where m h is the number of observations per hybrid m p is the number of observations per parent. This allocation minimizes the variance of a hybrid-parent contrast. Using a somewhat different optimality criterion, Dunnett (955) found the same optimal allocation. Note that complete romization would imply a single genotype per chip. By comparison, the optimal allocation (35) q implies that m h =m p ¼ ð ffiffiffiffiffi 3 ÿ Þ :464, where mh ¼ n m p ¼ n n, which is rather close but not equal q to ffiffiffiffiffi. The difference is mainly due to the incomplete blocking, with blocks corresponding to chips. DISCUSSION In this article we have derived formulas for the optimal allocation of resourses in cdna expression studies to reveal midparent heterosis or better-parent heterosis at the gene level. A common feature of both of these cases is that most of the resources are allocated to the parenthybrid pairs. The researcher needs to make up his mind as to which type of heterosis he wishes to assess. In the case of midparent heterosis, the parent-parent pair need not be tested at all, while with better-parent heterosis a small fraction of the total resources should be devoted to both parent-parent pairs hybrid-reciprocal pairs. We have not addressed the question of optimal sample size n. This may be determined by stard procedures (Steel Torrie 98). The sample size needed to detect heterosis will, among other things, depend on the variance. It should be stressed that variance will usually be gene specific, so optimal sample size will differ among genes. In designs with small sample sizes, efficient estimation of the variance is critical, it may be useful to borrow strength from other genes (Wright Simon 3), trading variance for some bias. As pointed out by a referee it may also be necessary to account for dye bias in variance estimation. The result that in optimal designs, parent-parent pairs provide no information regarding midparent heterosis contrasts, may seem trivial on first sight. It should be pointed out, however, that it does not generally hold in suboptimal designs is therefore not as trivial as it may seem. The reason is that parentparent pairs provide indirect information regarding heterosis contrasts. For example, data on the parent pair A-B on the parent-hybrid BA-A allow an indirect comparison for the pairing BA-B, since BA-A ÿ (A-B) ¼ BA-B. Therefore, it is often found (results not shown), that with suboptimal designs, the parent-parent pair provides information for the heterosis contrast. For optimal designs, this information vanishes in much the same way as information from indirect comparisons vanishes in a complete block design. In many experiments, the linear model needs to account for several fixed rom sources of variation, giving rise to a complex mixed linear model (Wolfinger et al. ). In this case, finding an optimal allocation will typically require numerical search strategies such as simulated annealing (Keller et al. 5). On the basis of the examples given in Keller et al. (5) it may be conjectured that the optimal allocation in more complex settings will not deviate dramatically from that derived in this article. To study heterosis, one may estimate the dominance ratio, u ¼ d/a (Kearsey Pooni 996). Using the d-method ( Johnson et al. 993) exploiting the fact that dominance additive gene effect estimates are stochastically independent, the approximate variance of the dominance ratio is dˆ var u varðdˆþ â d varðâþ a : ð36þ One might consider finding a design that minimizes this variance. This approach is not usually feasible, however, unless a priori information is available on both a d, which will rarely be the case. The same problem would apply if one were to work with the exact distribution of dˆ=â, assuming normality (Hinkley 969), or Fieller s (954) method (Piepho Emrich 5). Thus, it is preferable to optimize the design for contrasts related to either midparent heterosis or better-parent heterosis. I thank two anonymous referees for several helpful suggestions. LITERATURE CITED Birchler, J. A., D. L. Auger N. C. Riddle, 3 In search of the molecular basis of heterosis. Plant Cell 5: Dobbin, K., R. Simon, Comparison of microarray designs for class comparison class discovery. Bioinformatics 8: Dobbin, K., J. H. Shih R. Simon, 3a Questions answers on design of dual-label microarrays for identifying differentially expressed genes. J. Natl. Cancer Inst. 95: Dobbin, K., J. H. Shih R. Simon, 3b Statistical design of reverse dye microarrays. Bioinformatics 9: Dunnett, C. W., 955 A multiple comparison procedure for comparing several treatments with a control. J. Am. Stat. Assoc. 5: 96. Falconer, D. S., T. F. C. Mackay, 996 Introduction to Quantitative Genetics, Ed. 4. Longman, Harlow, UK. Fieller, E., 954 Some problems in interval estimation. J. R. Stat. Soc. B 6: Gibson, G., R. Riley-Berger, L.Harshman, A.Kopp, S.Vacha et al., 4 Extensive sex-specific nonadditivity of gene expression in Drosophila melanogaster. Genetics 67: Guo, M., M. A. Rupe, O. N. Danilevskaya, X. Yang Z. Hu, 3 Genome-wide mrna profiling reveals heterochronic allelic variation a new imprinted gene in hybrid maize endosperm. Plant J. 36: Harville, D. A., Matrix Algebra from a Statistician s Perspective. Springer, Berlin.

6 364 H.-P. Piepho Hinkley, D. V., 969 On the ratio of two correlated normal rom variables. Biometrika 56: (correction: Biometrika 57: 683). John, J. A., E. R. Williams, 995 Cyclic Computer Generated Designs. Chapman & Hall, London. Johnson, N. L., S. Kotz A. W. Kemp, 993 Univariate Discrete Distributions, Ed.. Wiley, New York. Kearsey, M., H. S. Pooni, 996 The Genetical Analysis of Quantitative Traits. Chapman & Hall, London. Keller, B., K. Emrich, N. Hoecker, M. Sauer, F. Hochholdinger et al., 5 Designing a microarray experiment to estimate dominance in maize (Zea mays L.). Theor. Appl. Genet. : Kerr, M. K., 3 Design considerations for efficient effective microarray studies. Biometrics 59: Kerr, M. K, G. A. Churchill, Experimental design for gene expression microarrays. Biostatistics : 83. Kollipara, K. P., I. N. Saab, R.D.Wych, M.J.Lauer G. W. Singletary, Expression profiling of reciprocal maize hybrids divergent for cold germination desiccation tolerance. Plant Physiol. 9: Lynch, M., B. Walsh, 998 Genetics the Analysis of Quantitative Traits. Sinauer, Sunderl, MA. Ni, N. Z., Q. Sun,Z.Liu,L.Wu X. Wang, Identification of a hybrid-specific expressed gene encoding novel RNA-binding protein in wheat seedling leaves using differential display of mrna. Mol. Gen. Genet. 63: Piepho, H. P., K. Emrich, 5 Simultaneous confidence intervals for two estimable functions their ratio under a linear model (in press). Searle, S. R., G. Casella C. E. McCulloch, 99 Variance Components. Wiley, New York. Simon, R., E. Korn, L. McShane, M. Rademacher, G. Wright et al., 3 Design Analysis of DNA Microarray Investigations. Springer, New York. Steel, R. G. D., J. H. Torrie, 98 Principles Procedures of Statistics: A Biometrical Approach. McGraw-Hill, New York. Stuber, C. W., S. E. Lincoln, D.W.Wolff, T.Helentjaris E. S. Ler, 99 Identification of genetic factors contributing to heterosis in a hybrid from two elite maize inbred lines using molecular markers. Genetics 3: Wolfinger, R., G. Gibson, E. D. Wolfinger, L. Bennett, H. Hamadeh et al., Assessing gene significance from cdna microarray expression data via mixed models. J. Comput. Biol. 8: Wright, G. W., R. M. Simon, 3 A rom variance model for detection of differential gene expression in small microarray experiments. Bioinformatics 9: Yang, X., K. Ye I. Hoeschele, Some E-optimal designs for cdna microarray experiments. ASA Proceedings of the Joint Statistical Meetings, New York, pp Communicating editor: J. B. Walsh APPENDIX We here derive Equation for matrix D. As we require n ¼ n 3 ¼ n 4 ¼ n 5 ¼ n, the matrix X9X is given by n n ÿn ÿn X 9X ÿn n n ÿn A ¼ A b b9 c ÿn ÿn n n 6 with ðaþ f ¼ðc ÿ b9a ÿ bþ ÿ : ða8þ To study the heterosis contrast (4), it is sufficient to find D. Using D ÿ ¼ A ÿ bb9c ÿ ¼ðn n ÞI ÿ n n 6 n n n n 6 n J ða9þ A ¼ðn n ÞI ÿ n J ; b ¼ÿn ; ðaþ ða3þ ðxi zj Þ ÿ ¼ x I ÿ z x z J ðaþ c ¼ n 6 n ; ða4þ where ¼ (, )9, I is a 3 identity matrix, J ¼ 9. Using results on the inverse of a partitioned matrix (Harville, p. 99) we find ðx 9X Þ ÿ ¼ D e e9 f ; ða5þ where D ¼ðAÿbb9c ÿ Þ ÿ ; e9 ¼ÿc ÿ b9d; ða6þ ða7þ (Searle et al., 99, p. 443), it can be shown that D ¼ I n n 6 n n n n n n n 6 n J : ðaþ Now the least-squares estimator dˆba ¼ k9 dðbaþ ðx 9X Þ ÿ X 9y has variance varðdˆba Þ¼k9 dðbaþ ðx 9X Þ ÿ k dðbaþ s D e ¼ k9 dðbaþ k dðbaþ e9 f ¼ 4 9Ds : ðaþ