Page 95 - Linear Models for the Prediction of Animal Breeding Values
P. 95

SETTING UP THE DESIGN MATRICES
        The X ′  and X ′  matrices, which relate sex effects for WWG and PWG, respectively, are:
              1     2
                 é 10 0 1 10ù                 é 000 11 0ù
            X¢  =  ê             ú  and  X¢ = ê               ú
              1
                                          2
                 ë 01 1 0 01     û            ë 0 1 1 000     û
                   é 30ù               é 20ù
            X¢ 1 X  =  ê  ú  and  X¢ 2 X  =   ú
                                     2 ê
                1 1
                   ë 03 û              ë 02 û
                           11
            In setting up X ′ R X , it is necessary to account for the fact that animals (one
                         1    1
        male and one female) have missing records for PWD. Thus:
                                            ⎡10 ⎤       ⎡20  ⎤  ⎡0.081 0.000 ⎤
                                11
                       11
                11
              ′
            XR X = r   m  W ′W +  r o  ′ B B = 0.025 ⎢  ⎥  + 0.028 ⎢  ⎥ ⎥  =  ⎢  ⎥
                   1
              1
                                            ⎣ 01 ⎦      ⎣ 02 ⎦  ⎣ 0.000 0.081 ⎦
        where the matrix W relates WWG records for animals 4 and 9 with missing records
        for PWD to sex effects and B relates WWG records for calves 5, 6, 7 and 8 to sex
        effects. The matrices W ¢  and B ¢  are:
                 ⎡10 ⎤           ⎡00 11    ⎤
            W ′ = ⎢  ⎥  and   ′ B = ⎢      ⎥
                 ⎣ 01 ⎦          ⎣ 11 00   ⎦
            However, all animals recorded for PWG also had records for WWG, therefore:
                                     é 20ù     é0.074 0     ù
              ¢
                22
            XR X   2  = r 22 X ¢ X 2  = 0.037 ê  ú  =  ê    ú ú
                       o
                          2
              2
                                     ë 02 û    ë 0     0.074 û
        and:
                                       ⎡ 20⎤     ⎡ −0.02  0.00⎤
              ′
                          ′
                12
            XR X = r o 12 X X = −  0.010 ⎢  ⎥  =  ⎢          ⎥
                   2
                          1
                             2
              1
                                       ⎣ 02 ⎦    ⎣  0.00  −0.02 ⎦
                                                            2
        Excluding ancestors, the matrix Z  is an identity matrix because every animal has a
                                      1
        record for WWG. Therefore, Z′ Z  = I and:
                                   1  1
                11
            Z′R Z  = diag(0.025, 0.028, 0.028, 0.028, 0.028, 0.025)
             1     1
        However:
            Z  = diag(0, 1, 1, 1, 1, 0)
             2
        indicating that calves 4 and 9 have no records for PWG, and:
                22
            Z′R Z  = diag(0.0, 0.037, 0.037, 0.037, 0.037, 0.0)
             2     2
        To account for ancestors (animals 1 to 3),  Z′R Z  and  Z′R Z  given above
                                                                  22
                                                     11
                                                   1    1       2    2
        augmented with three rows and columns of zeros.
            The other matrices in the MME can be calculated through matrix multiplica-
                                                   −1
                          −1
        tion. The matrix  A  can be set up and  A −1* G  (where  * means the Kronecker
        product) added to the appropriate matrices, as described in Section 5.2.2, to obtain
        the MME. The MME are too large to be presented but solutions from solving the
        equations are shown below, together with solutions from the univariate analyses of
        WWG and PWG.
        Multivariate Animal Models                                            79
   90   91   92   93   94   95   96   97   98   99   100