Page 139 - Becoming Metric Wise
P. 139

130   Becoming Metric-Wise


                     0 1
                       1
             Let D 5  @ A  and suppose that we want to calculate D D. First we
                       2
                                                                t
                       0
          note that this multiplication is mathematically possible as the number of
                       t
          columns of D is equal to the number of rows of D. The result of this
                                              0 1
                                                1

                                                2
          matrix multiplication is:  1  2  0    @ A  5 1 1 4 1 0 5 5 (this is a
                                                0
          (1,1)-matrix or a number, sometimes referred to as a scalar when matri-
          ces, vectors and numbers (scalars) are used in the same context).
                               t

             Note that also D D exists. This is a (3,3)-matrix obtained as follows:
          0 1                  0         1
             1                   1  2   0

             2
          @ A     1    2  0 5  @  2  4  0  A
             0                   0  0   0
             Using matrix multiplication we now prove four propositions showing
          how to derive the number of references of a given article, the number of
          articles written by a given author, the number of citations received by a
          given paper and the number of coauthors of a given paper, when a larger
          article-article matrix C or an author-article matrix W is given.

          Proposition 1: Given the citation matrix C, the number of references of
                            P  n                     t
          a given article d i is  j51 ij 5 C   Uð  Þ 5 C   C Þ , where U is the col-
                                 c
                                               ð
                                                      ii
                                            i
          umn vector completely consisting of 1’s.
             Proof: This result is easy to see since the C matrix consists of zeros
          and ones, ones if the corresponding cell is occupied and zero otherwise.
                                 P n
                                      c
          Keeping the row i fixed  j51 ij is just the number of ones in the i-th
          row. This is the number of times document d i has a reference, or the total
          number of references of document d i .
             Now C   U is an (m,1) matrix, i.e., a column vector. C   UÞ is the
                                                              ð
                                                                     i
          i-th   element   of  this  column    vector.  It   is  equal   to:
                    P  n         P n        P n
                                      c
                                                 c
                          c
          ð C   UÞ 5   j51 ij U j 5  j51 ij 1 5  j51 ij .
                 i
                            t
             Similarly C   C Þ 5  P n j51  ðÞ ij  ð  t  ji  P n j51 ij c ij 5  P n j51 ij . The last
                                      C : C Þ 5
                                                      c
                                                                 c
                      ð
                             ii
                                                    2
                                          2
          equality follows from the facts that 1 5 1 and 0 5 0.
             Next we consider an author-article matrix W.
             Assume that the (m,n)-matrix W, with elements w ij , is an author-
          article matrix. This means that w ij 5 1 if author i has authored (as sole
          author or as coauthor) article j, and zero if this is not the case. The
   134   135   136   137   138   139   140   141   142   143   144