Because the y-axis is Trump vote minus Republican vote and because the x-axis is Republican vote, there will always be a negative slope if you assume the Trump vote is less than 100% of the Republican vote. It would be a positive slope if the Trump vote is greater than 100% of the Republican vote. The only time it will be a zero slope line is of the Trump vote is exactly the same as the Republican vote.
It works like this:
- Suppose that Trump got 90% of the Republican vote.
- Suppose that a particular precinct is 100% Republican.
- In that precinct, Trump would get 90% of the vote.
- This would be a margin of -10% for Trump.
- Now suppose a precinct is 50% Republican.
- In that precinct, Trump would get 90% of the Republican vote, but that comes out to only 45% of the total vote.
- This would be a margin of -5%.
See the pattern? A negative slope is baked into the y-axis.
It's not normalized against the total vote but against straight party-line vote.
Re-watch the presentation.
So what is the equation for the graph? I thought it was Y = T - X, where T is the percentage of the total vote that went to Trump and X is the percentage of the total vote that went to other Republicans.
Watch his video from the 22 minute mark. You're wrong, there are two data points being compared. non-straight party republican voters and straight party republican voters. They should mirror each other to a degree, not deviate from the benchmark in a linear way. Here's the video just watch it for yourself:
https://www.youtube.com/watch?v=Ztu5Y5obWPk&t=3055s
It doesn't really matter what the subtracted percentage is. As long as you are comparing a difference in percentage with a percentage itself, you will get a slope unless the difference is near zero. All this analysis really proves is that Trump's share of the Republican vote is nearly a constant.
Does that mean the y-axis is the dem Trump vote?
its the non-straght-party ticket vote. I would have been one of these, because I didn't vote straight Rep but voted each candidate as I chose
So it’s like non-cis-gendered? Are you a fag?
BTW, purely for LULZ
Wish i was smart enough to chime in
no, you misunderstand the axes
if there were 1000 people, all registered republicans (not relevant), and 400 of them voted straight r ticket, and other 600 voted trump, this would be a dot at far right of x axis at 100% because among straight ticket votes R got 100% and dems 0%. then y axis dot would be way above the zero mark because trump got 50% more than the r straight ticket.
You're misunderstanding the graph.
In MI you can either vote straight ticket or individually.
The x axis is the percentage that chose republican of straight ticket voters.
The y axis is the difference between the percentage that votes Trump individually and those who voted straight ticket republican.
I won't say more because I don't want to help the dems.
But that's my point. If the y-axis is a difference in percentages and the x-axis is a percentage, the line will always be sloped unless the difference is very close to zero. As percentage increases, the difference increases. The only way to remove the slope is to normalize the y-axis by the straight ticket percentage.
So far every analysis I've seen of data fails to consider two important factors: failure to compensate for increased registrations in states where every registered voter was force fed a ballot and current 2020 data is not representative of complete population of voters (running calculations on incomplete data set?). Even the hacked data sets people are claiming to have can't be 100% accurate because even election centers don't have the data themselves or they'd have submitted it already. I'm working on a project similar but far deeper than Dr. Shiva's but I can't start it until 100% of precincts report. Sure I could use current data but if you look at maps many counties the reported percentages still to this date fluctuate from 87-98% of precincts reporting. Any report on those findings is conjecture at best because not only do you need 100% of it, but you need something to compare it to.
The problem I'm finding is that states are clearly planning to release the finalized counts on the day before the electors cast their vote. Theoretically we may only have 24hours between the release of data and votes being submitted. realistically it's mere hours to put together findings on the data and spread them to the public. I'm not sure that's something we should be gambling on. I'm not impressed with the math we've seen presented so far about vote counting patterns there's still so much to divulge.
It's not conjecture to use incomplete data.
Your model will just have a larger error variance. Even so, the error may be small.