Page 108 - Applied Probability
P. 108

5. Genetic Identity Coefficients
                              92
                                                                                              2
                                                                 2
                                   Next apply the inequality (a + b) ≥ 4ab to prove 4∆ 7 ≤ (4Φ ij ) ;
                                   finally, rearrange.)
                                 3. Calculate all nine condensed identity coefficients for the two inbred
                                   siblings 5 and 6 of Figure 5.1.
                                 4. The Cholesky decomposition of a positive definite matrix Ω is the
                                                                                          t
                                   unique lower triangular matrix L =(l ij ) satisfying Ω = LL and
                                   l ii > 0 for all i. Let Φ be the kinship matrix of a pedigree with n
                                   people numbered so that parents precede their children. The Cholesky
                                   decomposition L of Φ can be defined inductively one row at a time
                                   starting with row 1. Given that rows 1,... ,i − 1 have been defined
                                   and that i has parents r and s, define [4, 10]
                                                           0                j> i
                                                         
                                                         
                                                           1     1
                                                            l
                                                                 2
                                                 l ij  =   2 rj + l sj      j< i
                                                                   i−1   1
                                                                      2
                                                           (Φ ii −    l ) 2  j = i.
                                                                   k=1 ik
                                   Prove by induction that L is the Cholesky decomposition of Φ. Why
                                                                   1
                                                                                        1
                                                            1
                                                                                 1
                                   is l ii positive? (Hints: Φ ii > Φ ri + Φ si and Φ ij = Φ rj + Φ sj for
                                                            2      2             2      2
                                   j< i.)
                                 5. Explicit diagonalization of the kinship matrix Φ of a pedigree is an
                                   unsolved problem in general. In this problem we consider the special
                                   case of a nuclear family with n siblings. For convenience, number the
                                   parents 1 and 2 and the siblings 3,... ,n+2. Let e i be the vector with
                                   1 in position i and 0 elsewhere. Show that the kinship matrix Φ for the
                                                                                       1
                                   nuclear family has one eigenvector e 1 − e 2 with eigenvalue ; exactly
                                                                                       2
                                   n − 1 orthogonal eigenvectors  1    m−1  e j − e m, 4 ≤ m ≤ n +2,
                                                              m−3    j=3
                                                  1
                                   with eigenvalue ; and one eigenvector
                                                  4
                                                         4λ − 2
                                                e 1 + e 2 +    (e 3 + ··· + e n+2 )
                                                            n
                                   with eigenvalue λ for each of the two solutions of the quadratic equa-
                                   tion
                                                         1   n +1      1
                                                     2
                                                    λ − ( +       )λ +    =0.
                                                         2     4       8
                                   This accounts for n + 2 orthogonal eigenvectors and therefore diago-
                                   nalizes Φ.
                                 6. Continuing Problem 5, we can extract some of the eigenvectors and
                                   eigenvalues of a kinship matrix of a general pedigree [16]. Consider a
                                   set of individuals in the pedigree possessing the same inbreeding coef-
                                   ficient and the same kinship coefficients with other pedigree members.
                                   Typical cases are a set of siblings with no children and a married pair
   103   104   105   106   107   108   109   110   111   112   113