# What possibly can be the resolution?

#### zombie

##### New Member
Hi,

I am doing a research paper on social networking website. By calculating the number of likes, comments and shares on each of the politician' facebook page, i want to predict who might be the winner in the upcoming elections. Any suggestion which might be the relevant test. I thought of going ahead with t-test, but not sure as there will be data for more that 4 months?

#### JesperHP

##### TS Contributor
I would focus on defining the dependent variable. How many politicians are candidates? And how many observations do you have on this variable? Probably zero because they havn't won yet. Also considering historical elections the candidates are likely to be different unless defined as representative of parties. But facebook profiles are inherently personal. So you have no empirical possibilty to support the assumption of a connection between popularity on facebook and election winning. Hence you need **** good theoretical reasoning to suport this.

Consider the fact that opinionsurveys measuring asking what people intend to vote are quite uncertain up to at least a given period of time before the actual election ... some say they don't change the last couple of weeks ... anyway even these surveys are uncertain and prediction only possible from survey done within the last week before election. And this is in spite of the fact that these surveys actually ask the people who VOTE the theoretical argument being that ELECTIONS ARE DECIDED BY THOSE WHO VOTE.

I just don't see any reason fo assuming such a connection. What percentage of the people who vote check out the politicians facebook profile. And even if they do is it then the case, that their decision on what they vote is based on facebook. Why not commercials then.

The point is you will not be able to find the right test before you have given serious thought to your research question. In this case I would consider reformulating it. But offcourse I might be an ignorant mongrol maken yet another mistake.

#### zombie

##### New Member
I would focus on defining the dependent variable. How many politicians are candidates? And how many observations do you have on this variable? Probably zero because they havn't won yet. Also considering historical elections the candidates are likely to be different unless defined as representative of parties. But facebook profiles are inherently personal. So you have no empirical possibilty to support the assumption of a connection between popularity on facebook and election winning. Hence you need **** good theoretical reasoning to suport this.

Consider the fact that opinionsurveys measuring asking what people intend to vote are quite uncertain up to at least a given period of time before the actual election ... some say they don't change the last couple of weeks ... anyway even these surveys are uncertain and prediction only possible from survey done within the last week before election. And this is in spite of the fact that these surveys actually ask the people who VOTE the theoretical argument being that ELECTIONS ARE DECIDED BY THOSE WHO VOTE.

I just don't see any reason fo assuming such a connection. What percentage of the people who vote check out the politicians facebook profile. And even if they do is it then the case, that their decision on what they vote is based on facebook. Why not commercials then.

The point is you will not be able to find the right test before you have given serious thought to your research question. In this case I would consider reformulating it. But offcourse I might be an ignorant mongrol maken yet another mistake.
Hi,

I do apologise but i should have mentioned this earlier. I have date for 106 days (106 observation) for Barack Obama and Mitt Romney (U.S. Presidential Elections 2012) and the result is out (Barack Obama is the winner). The hypothesis is
i) The number of share, likes and comments (weightage for share is 40%, like is 30% and comment is 30%) can suggest who might be the winner.
ii) The percentage of votes secured by each candidate is proportional to the number of like, share and comment received on facebook page compared to the other candidate.

You are right that who will check facebook page before voting, but i think this gives an idea about the popularity of an candidate. Considering how many Americans are active on Facebook and how much attention is given to social marketing by politicians (i have seen dedicated team for facebook both by Barack Obama and Mitt Romney as every alternate day a picture, related to election campaign, is posted on Facebook).

Having said this, i would like to say that it has been rather difficult for everyone to suggest which will be an appropriate test and this is my last attempt and if i cant find a resolution i will leave this.