Page 167 - General Election
P. 167

While it is true that a 2% margin of error has been considered in the report but the
               confidence level considered is also far higher at 99.993665752% (4 sigma level) than
               considered by most commentators.





               Similarly, the question about what to use as the population for the study?

               To answer this question, it was first necessary to decide over the level at which the

               statistical guarantee was required. In this case the basic unit of the study was the
               EVM machine and it was required to check the working of the EVM through VVPAT
               counts. It is noteworthy that all EVMs are similar, having come up through the same
               manufacturing and testing process. Every machine is connected to a VVPAT. Before

               poll  each  EVM  undergoes  FLC  and  multiple  mock  polls  and  verifications  as  per
               established protocols in the presence of political parties. It is also to be noted that
               there  is  no  systematic  bias  in  allotting  particular  EVMs  to  particular  states  /

               constituency/ booths. Machines are randomized twice, first time for allocating to
               assembly constituencies and second time for allocating to polling stations.





               Thus, the report considered as a unit all the elections that are announced on a given
               day  and  typically  counted  on  the  same  day.  This  could  be  any  combination  of
               assembly elections in a few states, or Lok Sabha elections along with assembly

               elections and bye elections and so on. All were considered as one set of elections
               and since the Statistical unit of study was the EVM, the population consisted of all
               EVMs being used in a particular set of elections for which counting was done on the
               same day.


               However,  the  mandate  of  the  Hon’ble  Supreme  Court  to  the  ECI  to  tally  five
               randomly  chosen  VVPAT  systems  per  assembly  constituency  /segment  goes  far
               beyond the recommendation of taking a sample size of 479 as given by the expert

               committee.

               In fact, for the parliamentary elections, this mandate boils down to testing 20,625
               randomly chosen VVPAT systems (five in each of the 4125 assembly segments). If
               no mismatch in the EVM and VVPAT count is found in such a sample, then it can be

               said with very high level of confidence or with near certainty that the sanctity of
               the election process is not affected by the use of EVMs.




























                                                            165
   162   163   164   165   166   167   168   169   170   171   172