I will agree with that. As it seems that you do some research into this, I was curious if you may know a good way of detecting fraud at the county level. For instance, I was going to take a look at the historical correlations from past elections across lots of the counties and see if these correlations break down in this election. For instance, it seems there are a lot of interesting correlations that break down in this election. Florida and Georgia correlation is one of them. Upstate NY going more dem and Ohio/Iowa going more rep is another.
Just thinking out loud (may not work), one approach you might try to quantify the correlations in a useful way is to take a random subset of counties and use them as the explanatory variables in a logistic regression to make a prediction for a single target county. Repeat several times for the same target county with different random sets of counties used as explanatory variables so you get a whole bunch of predictions for the target county. If the actual vote for the target county is far outside of any of the predictions for that county, the target county is suspicious.
Just a follow-up on the regression: One would have to think a bit about how best to define the explanatory variables. For example, should they be raw percentages, or should they be the change in percentage compared to the prior year?
I will agree with that. As it seems that you do some research into this, I was curious if you may know a good way of detecting fraud at the county level. For instance, I was going to take a look at the historical correlations from past elections across lots of the counties and see if these correlations break down in this election. For instance, it seems there are a lot of interesting correlations that break down in this election. Florida and Georgia correlation is one of them. Upstate NY going more dem and Ohio/Iowa going more rep is another.
Just thinking out loud (may not work), one approach you might try to quantify the correlations in a useful way is to take a random subset of counties and use them as the explanatory variables in a logistic regression to make a prediction for a single target county. Repeat several times for the same target county with different random sets of counties used as explanatory variables so you get a whole bunch of predictions for the target county. If the actual vote for the target county is far outside of any of the predictions for that county, the target county is suspicious.
Just a follow-up on the regression: One would have to think a bit about how best to define the explanatory variables. For example, should they be raw percentages, or should they be the change in percentage compared to the prior year?