Page 329 - Contributed Paper Session (CPS) - Volume 6
P. 329

CPS1949 Khalid S.





               In this specification, although the weighting function grants more weight
            for the deprivation, the fact of not considering a logarithmic function allows
            to grant more importance for indices of deprivation translating less frequent
            symptoms of poverty. This approach tends to converge the weights of items
            measuring  the  deprivation.  Unlike  the  approach  of  Cerioli  and  Zani  which
            overrepresented the weight of the least wide-spread deprivation.
            iii) Weighting scheme of Betti & Verma
               This  weighting  scheme  is  based  on  two  principles:  i)  the  weight  to  be
            attributed measures the intensity of deprivation of an attribute; its value is an
            inverse  function  of  the  degree  of  deprivation  of  this  attribute  for  the
            population;  ii)  the  index  measuring  this  weight  aims  to  reduce  the  over-
            representation due to the risks of the high correlation between the attributes
            and to the redundancy of the information. So, the suggested function removes
            from the calculation items bringing the same information by eliminating their
            weights.









                depends on the distribution of the attribute j in the population, and
            depends on the correlation between xj and other attributes, it measures the
            average correlation of item j with all the other items. The more it increased the
            lower is the weight of the attribute j.     represents the level of correlation
            between two attributes j & m, I(.) is an indicator function and    is a pre-
            determined cutoff correlation level
            (in our case        ).
         iv)  The third weighting scheme combines the advantages of  Cerioli-Zani method
            and Betti-Verma method : It attributes an upper weight for the least wide-
            spread  deprivation,  and  limites  the  influence  of  the  correlation  and  the
            redundancy of the information on the weighting. Within the framework of this
            study, we adopted the following both specifications:



               This function would allow to attribute a weight to the item j by combining
            the methods of Betti & Verma and Cerioli & Zani.
               Similar to the specification BV-CZ, the second specification would allow us
            to adjust the function of the weight proposed by Desai-Shah by introducing


                                                               318 | I S I   W S C   2 0 1 9
   324   325   326   327   328   329   330   331   332   333   334