Correct approach to statistical analysis of voting results

manpreet Best Answer 2 years ago

Consider data consisting of voting results. Votes can be either individual votes of jury members or aggregated votes by certain communities (professional of geographical). In the end there is a set of data where each row has three fields: Voter, Candidate (the one who collects votes) and Place (or rank, voters put candidates on different places). There are several competitions, so candidates are measured multiple times.

Now we want ot analyze if some voters are biased, e.g. if they keep on giving some candidates certain places no matter if those candidates deserve it. I am not really sure what's the proper way of achieving this. Here is what I am trying to figure out:

1. In case votes are collected from communities, voting results are based on raw data of different sizes: one community can have 100K members while the other one has only 1000. Can we still assume that "places" represent ordinal data, not categorical?

2. What is the recommended technique to evaluate if a voter is biased? I can think of a simple measurement: an average difference between a candidate's place and the place he is given by a certain voter. The more the difference, the more the candidate is overrated by this voter. But I am not sure this is right thing to do because with such approach a single very low mark can balance multiple overrated high marks.

3. I've read about chi-square and Fisher's tests but not sure if it's applicable here. AFAIK they are applicable to categorical data.

Any help is appreciated.