Detailed methodology
A parliamentary constituency comprises of many small vidhan sabha constituencies. Total numbers of voters in a parliamentary constituency is essentially the sum of the voters in these vidhan sabha constituencies for example PCamethi= VStiloi+VSsalon+VSjagdishpur+VSgauriganj+VSamethi where PC=parliamentary constituency and VS= vidhan sabha constituency. We have to establish a mathematical relationship between the total votes that a party got in vidhan sabha election in the entire lok sabha constituency (which is essentially the sum of all the votes in various vidhan sabha constituencies which comprises the lok sabha constituency) and the total votes in the lok sabha election for the same parliamentary constituency. Such a relationship has to exist as same voters vote for the vidhan sabha election and the lok sabha election for a particular constituency. It may be week, but definitely some relationship has to exist. This relationship was found out by the following procedure which is regarded as polynomial regression analysis. * As in our case we see that the previous vidhan sabha election was held in the year 2012 and the lok sabha election is slated in the 2014 year. Hence we will historically pick those data sets where the lok sabha election was held after two years of the vidhan sabha elections. By establishing relationship between these data sets we assume that the same voting pattern will continue to occur also in the future. For example lets say that vidhan sabha election was held in the year 2004 and lok sabha election in 2006, den for constituency amethi we have
[congress votes(tiloi)+congress votes(salon)+congress votes(jagdishpur)+congress votes(gauriganj)+congress votes(amethi)]vidhan sabha 2004 should bear a mathematical relationship with congress votes(AMETHI)lok sabha 2006. * We will find the voters liking in each vidhan sabha and lok sabha election for example if in 2004 vidhan sabha election in the tiloi constituency if 44% was the total voters turn out and 45% of the votes casted were in favour of congress so voters liking = 45% of (44%) which is equal to 45/100*44. This figure qill be in %age similarly we will calculate voters liking of salon,jagdishpur,gauriganj and amethi for vidhan sabha election of 2004. Next we calculate voters liking for parliamentary election of 2006. Sum of voters liking for different vidhan sabha constituency in vidhan sabha election of 2004 is plotted on X axis and voters liking for lok sabha constituency is plotted on Y axis. Like wise voter liking for other year set where the time gap between the vidhan sabha and lok sabha election is two yyears is plotted on the respective axis. * After the data has been plotted we try to plot a curve which bests represent the relationship between all the points. This is called polynomial regression analysis we do this with help of software and equations of the various degree are produced as a result of software analysis
An equation of fourth degree would look like ax^4+bx^3+cx^2+dx+e. Values of the constant a,b,c,d,e are calculated by the computer software. Hence by substituting the value vidhan sabha election liking of 2012 in such a equation we get the expected voters liking for the year 2014 lok sabha elections. This is the base value. * We c dat as we go on decreasing the time lag between the lok sabha elections and vidhan sabha elections, less is the error of the regression formula. For example if both elections are held in the same year than the lok sabha election results can be predicted with the help of vidhan sabha data by the regression analysis i.e. the formula obtain from the software. But as the time lag goes on increasing, more unpredictable the result of the lok sabha becomes and it cannot be accurately predicted by the regression analysis alone. Hence we use some other method..it can be said that in situation where both election occurs in the same year then the results of vidhan sabha elections predict lok sabha results with almost certainity by the regression analysis.