no, I mean what exactly was he plotting

Ask him ? https://vashiva.com/contact/

I doubt you'll see a response but I feel like anyone trying to prove or disprove something should be willing to share applicable data to make the results irrefutable.

He plotted exact numbers of ballots checked as Party Ticket, and exact numbers of non-party ticket per precinct. He then plotted percentage of party ticket republican per precinct on the x axis, and percentage of individual votes for trump on the y axis.

yeah that's what I thought and tried to model ... point is if just a small percentage of Dems broke with straight party vote and voted for Trump you can get something like what he plotted

It's not that the some percentage broke. It's that the percentage that broke increased with how republican the precinct was. But also that it didn't start increasing until the precinct was 25% Republican

yeah I'm not buying his analysis and with a simple model and spreadsheet I can get a curve that does something like what the data shows

x-axis is percentage Republican so to the left is heavy Democrat, if just a small percentage of Democrats break with party and vote Trump then on the left of the graph you get more Trump votes than downline Republican ... then if a fraction of Republican vote party downline but NOT Trump as you move right towards heavier Republican precincts you less Trump votes than downline Republican ... adjust the relative percentages and you can put the break over point anywhere you want

I'm looking at a similar pattern for the right side of his graph in GA

They are comparing the fraction, not the total amount

Right, that's what I did

So it's an increasing fraction of total voters yes?

not sure what you're asking but the point is that the horizontal axis is percent Republican, so as you move left there are more Democrats voting and less Democrats voting as you move right ... what he observed was a downward trend of Trump votes divided by Republican downline votes AND what he said was if say Romney's analysis is correct that some Republicans voted party downline but NOT for Trump that you would have a flat curve under the 1.0 ratio line but what Dr. Shiva failed to account for is that even if a small percentage of Democrats break from straight party and vote Trump then you can get a curve or trend line similar to what he observed

When you say % republican, do you mean % voted party ticket republican, or % registered republicans?

Finally, I'm not buying he claim that the shape of the trend line is evidence of election fraud ... he claims it's evidence of a computer algorithm but I claim it could be just evidence of voter preferences within a particular county

regarding % republican I assume "voted" because I believe that's all the data he had (I assume the ballot data is secret and thus you won't know if the vote cast was registered Republican or not)

Yeah, agreed

okay then back to the court cases, I found this video useful to summarize the court cases currently in play https://youtu.be/h6pI3-nWmSQ

Still the down ballot comparisons are relevant cross states

https://bitcadia.github.io/DownBallot/Outputs/CountyDownBallotDiffsByStraightCompare.html

to what end? to detect possible voter fraud? I don't think so

> Also I think his recommendations about more transparency and greater access to data and the ability to audit are all good things
@meglide

If 100% of Michigan counties matched this trend, you might be onto something--that the more Republican a county was, the more the voters disfavored Trump. However, if this pattern exists in only a few counties, then we are looking at a serious problem.

I have not seen a further analysis of more Michigan counties, but if you compare Oakland and Wayne counties, which are neighboring counties, the difference in the distribution is absolutely stark. These were both Democrat counties. Is there something about the county line that would make Republicans less likely to support Trump? I don't think so. It seems clear that some manipulation was happening in Oakland county that wasn't happening in Wayne.

> If 100% of Michigan counties matched this trend, you might be onto something--that the more Republican a county was, the more the voters disfavored Trump. However, if this pattern exists in only a few counties, then we are looking at a serious problem.

why? there are any number things which might explain this other than voter fraud'

for example, suppose there is a Republican candidate that was liked by voters in the county but didn't get along with Trump and/or some of Trump positions ... if John McCain was still alive you could see folks in AZ voting for him but not for Trump

another example might be that some business in that county had to lay workers off due to say the tariffs that Trump placed on China, again the Republicans might like their local candidates and not like Trump

Now I don't know anything about these counties because I don't live anywhere near MI and never have ... BUT I don't consider this particular voting pattern evidence of fraud and I don't think a court would either ... Dr. Shiva seemed to suggest you could never get this pattern and was able to simulate this by making a couple simple assumptions about human behavior in the county, so mathematically it is certainly possible

@meglide for an individual person or county, that might be true. As datasets get larger, however, they conform to the same overall 'shapes' if they are not being manipulated.

I don't think this proves fraud. It does prove that we need to dig into the details, double-check and make sure there was no manipulation.

In particular, I'd like to see a county-by-county analysis of every county in Michigan, and compare them all, and then compare systems in use between counties that conform to a natural distribution to the ones that don't.

yes and a county-to-county comparison showed NO such pattern

I think you've re-made my point for me

I may have missed that analysis, can you post?

@DrSammyD did the analysis and has posted it twice

@meglide I did see a link but it's a 404

@meglide Also, it appeared from comments that this was a county count, not a precinct count as Dr. Shiva was doing. It's critical to the analysis that precinct-level data is used.

> @meglide https://bitcadia.github.io/DownBallot/index.html

works for me

OK, that one works... again I see no precinct level data. Also the quirk that is enabling this analysis is unique to Michigan or at least isn't shared by every state.

https://youtu.be/TTw4aErwego

I think what may be somewhat unique to MI is that they allow you to vote straight party by checking just one box (or filling in just one oval) whereas this is option is not available in most other states

I agree, which is why this analysis is critical. If there's no manipulation, further analysis will show that. Statistics isn't useful for proving anything, it useful in showing you where to look for evidence.

yes precisely ... and I took issue with Dr. Shiva's assertion that the particular pattern he was seeing had to be the result of fraud speculating that it was a computer algorithm, whereas I claim just some simple assumptions about voter patterns in that county could produce that same pattern ... as why that voter pattern should exist and what motivates humans to vote that way I don't know, my interest was that the mathematics allow for it and so you're now in the realm of trying to explain human behavior