Shiva says that the trend line should be horizontal, but that didn't make sense to me based on the way the y axis is plotted, so I created a scatter plot from random numbers using the same math, and it has an obvious downward trend.
There might be other oddities in the actual election graph, but the trend line slope is not odd, it's exactly what you would expect from a random distribution.
He plotted the % a precinct leans R against how much it voted for trump (relative to the baseline expectation that R's vote trump) and observed that the more heavily it leaned R's the lower the Trump vote was relative to expectations.
You are plotting, what? Independents vs republicans based on a random distribution? I don't understand what you're doing. Maybe you're smarter than me. Email Shiva if you think you're analysis is correct, I bet he'd answer.
Since the two populations are independent and the measure is percentage, it's not far fetched. The population of non RSP could include Rs who dislike one or more down ballots, but are "mostly" or even partly Republican. There are also true independents and cross party voters from the Ds who didn't go DSP and deviated at least on Trump. I don't think you would see an arbitrary linear depression. In general, a downward trend sounds plausible, but so do unobserved outliers that ARE observed in the low ranges as s completely different trend. I think artificial manipulation is viable (too pretty, slope is fsirly consistent across different counties despite varying x intercepts). We don't see any outliers.
Contrived example: deep red 90% RSP. Only 1 person in non RSP and they like Trump. He beats RSP performance. The smaller the pop, the fewer he needs to pick up to match or beat RSP perf.
For that to happen, all the Biden-Hating-Trump-Loving Democrats that blow out the candidate vote 90% to 10% would have to live in deep red districts where the party vote is also 90% to 10%, and that sounds totally far-fetched to me.
This characterization ignores Trump Rs who don't vote straight ticket or abstain from some down ballot (like me in at least 1 election). The performance over RSP is sensitive to this population, the composition of which may feasibly mirror the larger RSP (if you're living in a deep red district, chances are you probably identify with your neighbors to some extent absent die hard contrarians).
I'm offering AN explanation of why the graph looks the way it does, not necessarily THE explanation.
My characterization certainly ignores other possibilities, but my characterization follows the data.
Shiva's characterization is that, "if Trump wins the Party Vote by a landslide, then he should also win the Candidate vote by a landslide". Further, he claims that since the data doesn't meet his pre-concieved expectations, then there must be fraud involved, followed by, "Trust us, we're experts, we know these things."
He doesn't show us what a random distribution should look like, and he doesn't show us a REAL election result KNOWN to be WITHOUT fraud, to see what the distribution should look like. He shows us chart after chart with the same result and concludes, "See, it's all fraud."
To your point. It is feasible that Candidate voting could track with the Party Vote, but the data should bear that out, and if it doesn't bear that out, we can't simply conclude that it's fraud, and we should be especially wary of simply believing a claim because it satisfies our preconceived notions or fits our narrative when it defies the data.
There are numerous reasons why people vote candidate by candidate. They could be moderates, they could be Republicans voting for Biden or Democrats voting for Trump, or Republicans and Democrats simply choosing to vote candidate by candidate. The more variables at play, the less likely I am to expect a direct correlation between this group of voters, which could be very diverse, and the Party Voters, who are much more politically homogeneous.
"if Trump wins the Party Vote by a landslide, then he should also win the Candidate vote by a landslide"
I don't recall him making that exact claim, although it was an hour long video.
He doesn't show us what a random distribution should look like, and he doesn't show us a REAL election result KNOWN to be WITHOUT fraud, to see what the distribution should look like.
Wayne county is a counterpoint that supports the (assumed) synthetic examples he portrays in the methodology section for under/over performance (all with the proposed horizontal curves). Wayne county seems to defy expectations the other counties present (at least for the digital voting stuff).
but the data should bear that out
Wayne County is substantially different. That's data. The commonality between the three counties presented could be an entirely natural phenomenon; I could also have a special rock in my pocket that keeps tigers away because there are no tigers around me.
we can't simply conclude that it's fraud
I don't think they did. I think there was a reasonable discussion toward the end with respect toward further questions that have been raised, and where to investigate. The underlying argument is that linear trends are fairly artificial in natural signals. The lack of outliers is another red flag. The criticism I have of his explanation (particularly the vote-switch portion where he walks the data) is showing the jump between adjacent precincts and identifying 200 votes "going to Biden." I didn't find that convincing, partly due to momentary ignorance. Still mulling it over (need to rewatch).
The more variables at play, the less likely I am to expect a direct correlation between this group of voters, which could be very diverse, and the Party Voters, who are much more politically homogeneous.
I think that is exactly what they are saying. If there is such probably diversity, why is there a direct negative correlation across three different counties with hundreds of precincts (aside from Wayne)? Where's the diversity, why are there are no exceptions to the rule? (outliers).
Good explanation. We would expect odds of finding favorable converts higher in districts with higher populations (of straight ticket R), since part of that convert population (non-straight ticket) would be majority ticket R voters who deviated or abstained somewhere down ballot, as well as the mash of independents and non-straight ticket D converts. The expectation (as with Wayne county) is that republican straight ticket percentage and non-repub straight ticket for Trump percentage are positively correlated (no idea how strong though). The positive correlation would offset the negative clean slope we see and at worst make it shallow, at best keep it horizontal (with outliers). My biggest hangup is the lack of clusters/outliers, and the comparison with Wayne County. Interesting analysis.
Jury is out of whether Shiva's stuff is "proof," but I see it as a valid red flag indicating unnatural behavior; and one that is consistent across counties where machine tabulation is used.
I don’t care what the data is. If you take Y-X over X and chart it on the above chart all points at X0 have to be at or above Y0 and all points at x100 have to be at or under Y0.
Let's say the candidate vote is always 50%. The higher the Republican Straight Party Vote, the more you are subtracting from the 50% candidate vote, the lower the Y-Axis. When X is 50%, Y is 0%. When X is 70%, Y is -20%. When X is 100%, Y is -50%.
The only manipulation here is the math for the Y-Axis.
There are some assumptions about his process, but it's not so easily discredited. Your counter-argument assumes that in a heavy red district, the voters will be predominantly voting the Republican ticket, as opposed to picking each candidate individually. I'm not so sure that party voting is that popular, even in districts which lean heavily.
And even then, such a strong and consistent downward correlation is unlikely. I would still expect a few outliers in the split areas (50% on x-axis) which would go for Trump above the norm. But NONE existed.
I'm not saying there's no fraud, but shiva made certain specific claims that don't add up, especially about a normal distribution having a horizontal trend line.
A random distribution would absolutely trend down. The math for the y axis includes -x. The bigger x, the more it's subtracting from y.
The whole presentation hinges on the assumption that a straight party Trump blow out must be a more conservative district. That could be a wrong assumption, but even if only a single person votes straight party for Trump, that's still 100%, which Is an impossible number for the candidate vote to overcome.
yes it slopes down because the data slopes down. the point is that the data shouldnt slope down.
I thought you were looking at Dr Shiva's charts https://www.bitchute.com/video/XAmigqBh8zAK/ but that is not what this is. What is this?
Shiva says that the trend line should be horizontal, but that didn't make sense to me based on the way the y axis is plotted, so I created a scatter plot from random numbers using the same math, and it has an obvious downward trend.
There might be other oddities in the actual election graph, but the trend line slope is not odd, it's exactly what you would expect from a random distribution.
He plotted the % a precinct leans R against how much it voted for trump (relative to the baseline expectation that R's vote trump) and observed that the more heavily it leaned R's the lower the Trump vote was relative to expectations.
You are plotting, what? Independents vs republicans based on a random distribution? I don't understand what you're doing. Maybe you're smarter than me. Email Shiva if you think you're analysis is correct, I bet he'd answer.
This is nonsense
Yeah the so-called Inventor of Email is a charlatan.
Believe the science cough when it’s politically convenient cough!!
Why only one trend? Empirical data was definitely not sloping and flat until the magic 25%ish zone.
Since the two populations are independent and the measure is percentage, it's not far fetched. The population of non RSP could include Rs who dislike one or more down ballots, but are "mostly" or even partly Republican. There are also true independents and cross party voters from the Ds who didn't go DSP and deviated at least on Trump. I don't think you would see an arbitrary linear depression. In general, a downward trend sounds plausible, but so do unobserved outliers that ARE observed in the low ranges as s completely different trend. I think artificial manipulation is viable (too pretty, slope is fsirly consistent across different counties despite varying x intercepts). We don't see any outliers.
Contrived example: deep red 90% RSP. Only 1 person in non RSP and they like Trump. He beats RSP performance. The smaller the pop, the fewer he needs to pick up to match or beat RSP perf.
This characterization ignores Trump Rs who don't vote straight ticket or abstain from some down ballot (like me in at least 1 election). The performance over RSP is sensitive to this population, the composition of which may feasibly mirror the larger RSP (if you're living in a deep red district, chances are you probably identify with your neighbors to some extent absent die hard contrarians).
I'm offering AN explanation of why the graph looks the way it does, not necessarily THE explanation.
My characterization certainly ignores other possibilities, but my characterization follows the data.
Shiva's characterization is that, "if Trump wins the Party Vote by a landslide, then he should also win the Candidate vote by a landslide". Further, he claims that since the data doesn't meet his pre-concieved expectations, then there must be fraud involved, followed by, "Trust us, we're experts, we know these things."
He doesn't show us what a random distribution should look like, and he doesn't show us a REAL election result KNOWN to be WITHOUT fraud, to see what the distribution should look like. He shows us chart after chart with the same result and concludes, "See, it's all fraud."
To your point. It is feasible that Candidate voting could track with the Party Vote, but the data should bear that out, and if it doesn't bear that out, we can't simply conclude that it's fraud, and we should be especially wary of simply believing a claim because it satisfies our preconceived notions or fits our narrative when it defies the data.
There are numerous reasons why people vote candidate by candidate. They could be moderates, they could be Republicans voting for Biden or Democrats voting for Trump, or Republicans and Democrats simply choosing to vote candidate by candidate. The more variables at play, the less likely I am to expect a direct correlation between this group of voters, which could be very diverse, and the Party Voters, who are much more politically homogeneous.
I don't recall him making that exact claim, although it was an hour long video.
Wayne county is a counterpoint that supports the (assumed) synthetic examples he portrays in the methodology section for under/over performance (all with the proposed horizontal curves). Wayne county seems to defy expectations the other counties present (at least for the digital voting stuff).
Wayne County is substantially different. That's data. The commonality between the three counties presented could be an entirely natural phenomenon; I could also have a special rock in my pocket that keeps tigers away because there are no tigers around me.
I don't think they did. I think there was a reasonable discussion toward the end with respect toward further questions that have been raised, and where to investigate. The underlying argument is that linear trends are fairly artificial in natural signals. The lack of outliers is another red flag. The criticism I have of his explanation (particularly the vote-switch portion where he walks the data) is showing the jump between adjacent precincts and identifying 200 votes "going to Biden." I didn't find that convincing, partly due to momentary ignorance. Still mulling it over (need to rewatch).
I think that is exactly what they are saying. If there is such probably diversity, why is there a direct negative correlation across three different counties with hundreds of precincts (aside from Wayne)? Where's the diversity, why are there are no exceptions to the rule? (outliers).
Good explanation. We would expect odds of finding favorable converts higher in districts with higher populations (of straight ticket R), since part of that convert population (non-straight ticket) would be majority ticket R voters who deviated or abstained somewhere down ballot, as well as the mash of independents and non-straight ticket D converts. The expectation (as with Wayne county) is that republican straight ticket percentage and non-repub straight ticket for Trump percentage are positively correlated (no idea how strong though). The positive correlation would offset the negative clean slope we see and at worst make it shallow, at best keep it horizontal (with outliers). My biggest hangup is the lack of clusters/outliers, and the comparison with Wayne County. Interesting analysis.
Climate science skeptic stats guy [Stephen McIntyre] (https://twitter.com/ClimateAudit/status/1326897315324567553) has a thread looking at some of this from a different perspective and tries to include D data as well.
Jury is out of whether Shiva's stuff is "proof," but I see it as a valid red flag indicating unnatural behavior; and one that is consistent across counties where machine tabulation is used.
If dems are only "just" trying to beat Trumps #'s then wouldn't their #'s trend downwards anyway?
They only needed enough votes to cover Trumps. So as Trumps #'s trended down they wouldn't need as many #'s to compensate.
If at 12am Trump has 100,000 votes, they only need 100,001 At 1am Trump has 90,000, they need 90,001...
Right?
source?
It has to slope down. As all points at x0 have to be above y0 and all points at x100 have to be below y0.
I don’t care what the data is. If you take Y-X over X and chart it on the above chart all points at X0 have to be at or above Y0 and all points at x100 have to be at or under Y0.
But it still has to slope down.
X-Axis = % straight Republican Votes Y-Axis - % candidate votes - % straight Republican votes
Let's say the candidate vote is always 50%. The higher the Republican Straight Party Vote, the more you are subtracting from the 50% candidate vote, the lower the Y-Axis. When X is 50%, Y is 0%. When X is 70%, Y is -20%. When X is 100%, Y is -50%.
The only manipulation here is the math for the Y-Axis.
Now new smart man say something differnt
Not sure what the joke or real argument is supposed to be.
There are some assumptions about his process, but it's not so easily discredited. Your counter-argument assumes that in a heavy red district, the voters will be predominantly voting the Republican ticket, as opposed to picking each candidate individually. I'm not so sure that party voting is that popular, even in districts which lean heavily.
And even then, such a strong and consistent downward correlation is unlikely. I would still expect a few outliers in the split areas (50% on x-axis) which would go for Trump above the norm. But NONE existed.
I'm not saying there's no fraud, but shiva made certain specific claims that don't add up, especially about a normal distribution having a horizontal trend line.
A random distribution would absolutely trend down. The math for the y axis includes -x. The bigger x, the more it's subtracting from y.
The whole presentation hinges on the assumption that a straight party Trump blow out must be a more conservative district. That could be a wrong assumption, but even if only a single person votes straight party for Trump, that's still 100%, which Is an impossible number for the candidate vote to overcome.
So convinced