Using Math, a simple way to double check a polling companies information. see if it makes sense
I constantly find myself trying to analyze things until i just get myself too confused, then come back and try to simplify, so when I speak It is usually only after I feel I know the data so well that I can start making general assumptions, or hypothisis than I can use to try to recreate data, and even go back to previous years data and see if my theory still holds true.
Example I have a theory that you can take an over sample and the spread of independent voters and make a prediction. This is mainly achievable right now because the republicans and democrats are so polarized that they almost cancel each other out, or in other words 95% Dems are voting for Obama and 95% of Rep are voting for Romney... So here is my over simplified formula (Overskew + Spread of Independents divided by two = poll spread)
I tried it on 2008 presidential election where independents were Obama +8 and oversampling was Dem+7. (8+7)/2 = 7.5%, and that was the spread of that election. Usered it in averages from moth ago independents +0 and oversampling of Dem+8 (8+0)/2 =4.0% and that was Obama's average lead about a month ago. I try it today where independents -8 and over sampling of Dem+4 (4-8)/2 = -2.0% or Romney up 2% all pretty reasonable
This can be done with men/women, ethnic groups, absentee and early voting vs election day voting compared to 08 04 00 etc, but this has been the easiest I could simplify with some correctness.
Lastly this enables me to calculate individual polls for its true accuracy, for example when someone comes out and says in dependents up 20% for Romney but polling says it is 2 point spread, it screams error, or if men are +10% and women suddenly even, this also screams error because math is not working.....
Example of error is PPP poll today that said independents +16 will an oversample od Dem+4 and they say Romney up 1%. (4-16)/2 = 6% which infers that Romney has a lead of 6%/ So either my assumptions are wrong, or PPP is misrepresenting something...NOTE I have other examples of PPP misrepresenting information I can share if you wish to know...
This would readjust the below to CNN O+4, Purple O+2.5, ARG R+5, Ras O+ 0.5, Time R+3, USA R +0.5, Suffolk R+8, PPP O +0.5, CBS O +1, Gravis R+5, Fox R+8 with an avg of R+2
OHIO
