Unfortunately, Dr. shiva's analysis is also flawed. The y-axis needs to be normalized by Repub straight ticket vote for any slope in the curve to be significant. Otherwise, you're just showing that differences in raw percentages grow with the magnitude of the raw percentages, which is self-evident and not a sign of fraud.
t = T/R = ratio of Trump vote to straight ticket Repub
T can therefore also be described by the equation: T = tR
Therefore, the y-axis of Dr. Shiva's graph can be written as: T - R = tR - R = (t - 1)R
In other words, the slope of the graph (t - 1) solely depends on the ratio of Trump vote to straight ticket Repub vote. If that ratio is greater than 1, it will have a positive slope. If that ratio is less than 1, it will have a negative slope. The only time there will be no slope is if that ratio is equal to 1.
I think the point was 1) that the curve unnaturally makes a right hand turn (the D district he showed also had a slope he didn't call that out). But more importantly, that 2) the slope is disconcertingly steep.
The maximum negative slope such a line can have (with a constant t) is -1, though. That means t=0, and no one voted for Trump. I eyeballed the slope in the video: Oakland county had (.2, .5) and (.6, -.19), that gives us a slope of -.775, which means t=.225. Kent county had (.25, .5) and (.77, -.25), a slope of -1.44 and a corresponding t=-0.44.
Caveat Emptor here is that I don't buy the analysis that comes up with tens of thousands of missing votes unless there's some extra-ordinary evidence-- I think the more likely answer is that there's a good real-world explanation. But, on its face, looking at districts that voted 70% for Republicans, but 40% for Trump sounds fishy.
Unfortunately, Dr. shiva's analysis is also flawed. The y-axis needs to be normalized by Repub straight ticket vote for any slope in the curve to be significant. Otherwise, you're just showing that differences in raw percentages grow with the magnitude of the raw percentages, which is self-evident and not a sign of fraud.
wrong
I can prove it with algebra if you like.
T = Trump vote pct of total
R = straight ticket Repub vote pct of total vote
t = T/R = ratio of Trump vote to straight ticket Repub
T can therefore also be described by the equation: T = tR
Therefore, the y-axis of Dr. Shiva's graph can be written as: T - R = tR - R = (t - 1)R
In other words, the slope of the graph (t - 1) solely depends on the ratio of Trump vote to straight ticket Repub vote. If that ratio is greater than 1, it will have a positive slope. If that ratio is less than 1, it will have a negative slope. The only time there will be no slope is if that ratio is equal to 1.
I think the point was 1) that the curve unnaturally makes a right hand turn (the D district he showed also had a slope he didn't call that out). But more importantly, that 2) the slope is disconcertingly steep.
The maximum negative slope such a line can have (with a constant t) is -1, though. That means t=0, and no one voted for Trump. I eyeballed the slope in the video: Oakland county had (.2, .5) and (.6, -.19), that gives us a slope of -.775, which means t=.225. Kent county had (.25, .5) and (.77, -.25), a slope of -1.44 and a corresponding t=-0.44.
Caveat Emptor here is that I don't buy the analysis that comes up with tens of thousands of missing votes unless there's some extra-ordinary evidence-- I think the more likely answer is that there's a good real-world explanation. But, on its face, looking at districts that voted 70% for Republicans, but 40% for Trump sounds fishy.