Page 223 - Becoming Metric Wise
P. 223

214   Becoming Metric-Wise


          when citation scores are not natural numbers then using this variant is, in
          our opinion, the method of choice. It is defined as follows:
             Let P(r) denote the magnitude value of the rth source and let P(x)denote
          the piecewise linear interpolation of the sequence (r, P(r)), r5 1, 2, .. .,then
          h r is the abscissa of the intersection of the function P(x) and the straight line
          y 5 x (if this intersection exists). Its value is then given by:

                                   ðh 1 1ÞPðhÞ 2 hPðh 1 1Þ
                             h int 5                                   (7.9)
                                     1 2 Pðh 1 1Þ 1 PðhÞ
             Clearly h int $ h with equality sign only if h 5 P(h).
             Considering the previous examples we have:
             For [7,4,3,1,0] h 5 3, P(3) 5 3; P(4) 5 1 and h int 5 (4 3 3    1)/

          (1 2 1 1 3) 5 9/3 5 3. As long as the third publication receives exactly
          3 citations h int stays equal to 3. For [6,5,4,3,2,1,0] we find
          h int 5 (4 4 2 3 3)/(1 2 3 1 4) 5 7/2 5 3.5. For [10, 2, 2, 2] h int is clearly



          2. For [10, 8, 1, 0] it is (3 8 2 2 1)/(1 2 1 1 8) 5 22/8 5 2.75. In most

          cases, but not always, h int , h rat , see (Guns & Rousseau, 2009a).
             It was mentioned that the interpolated h-index could also be used
          when citation values are not necessarily natural numbers, which happens
          e.g., when using fractional counting. It is then possible that the most-
          cited citation of a set of publications has a value of 0.5, in which case the
          line y 5 x does not intersect P(x). This explains the restriction in the
          definition (the part “if the intersection exists”). A solution for this prob-
          lem has been suggested in (Rousseau, 2014c).


          7.4.3 The m-quotient
          It has been mentioned that in its original version the h-index favors
          scientists with a longer career. Hirsch (2005) provided a solution for this,
          namely by dividing the standard h-index by the length of a scientist’s
          career expressed in years. We note that there is an unspoken assumption
          used in this proposal. This assumption is that in the “regular” case an
          h-index grows linearly over the years. If this regular growth were curved
          then either younger or older scientists would be favored by this correc-
          tion. It has been observed that by and large the h-index of a scientist
          grows linearly, but, of course, there are many exceptions. Surely, the
          higher the h-index the more citations are needed (we remarked earlier
          that this number can be as high as (2h 1 1)) to increase it by one unit. As
          many articles in the h-core tend to be older ones, for which it is less
   218   219   220   221   222   223   224   225   226   227   228