I did. I went to the county website and followed the link to results, which was hosted by clarity elections. I suspect a number of counties are on that same site but I could not find a way to browse it. Otherwise I guess have to go to each county. And the ingest might need to be different for each.
I'm going to try to get more counties but if you find any data I'd love to add it. I don't know if you've used git but you can add stuff to your copy of the repository and make what's called a pull request that allows me to import it into the main hosted copy. Or if you can somehow post it somewhere I can d/l and add it.
Yes, I plan to try to collect more datasets in that github (I also have to do some work on the ingest scripts). I guess my first question is whether we can believe that plot or is there something wrong again.
The analysis was:
X axis is (straight_r_count+split_r_count) / total_votes -- the fraction of all votes that are for trump
Y axis is straight_r_count / total_votes - split_r_count / total_votes -- the difference in fractions between straight ticket and split ticket
This metric appears to have an extremely linear trend.. much more so than Shiva's original result! If you have time can you possibly check my work?
I see what you mean. My guess is that if you do split or mixed, it automatically assigns the votes to all the individual race tallies and does not tally in the straight party sheet.
I don't have any first hand knowledge, but another comment on this post said they observed the machine rejecting ballots that selected both a straight party and another down ballot. So maybe it's not properly implemented everywhere.
In any case, the data they publish doesn't account for it, so if it is allowed it must be compensated the way I describe or else there would be double counts.
This analysis depends on separate counts for straight ticket, is that the case for co?
My understanding was that you can't vote straight with exceptions. You can either vote straight or choose. So the data has 1 sheet for straight ticket votes and a second sheet for selection for president but those are mutually exclusive
That's easy enough. Here is the plot for your suggestion:
https://github.com/vermiculita/expert-carnival/blob/master/tmp/fig4.png
Note also my plot that I described above is very similar to yours except the X axis is straight+split as the numerator in the X axis, rather than straight only:
https://github.com/vermiculita/expert-carnival/blob/master/tmp/fig3.png
That shows a very strong linear trend. BUt maybe Im not thinking through whether the measures are independent.
One thing that I will try tomorrow is run the same analysis for biden / D votes. I have to re-pull the data cols out of the csvs so it's a little bit more work.
Sure.
I think his intent was to formulate a proxy for republican sentiment and then compare the actual votes in these two categories (straight and split ballots) to that proxy.
The proxy is the fraction of straight ballots voted as R. So that is the X axis, and the Y axis is the difference between the rates of straight ballots voting R and split ballots voting trump.
That's how I understand his graph.
This would work if the two measures were independent, but they aren't, because as you pointed out, among the set of votes for Trump, if the voter doesnt split the ballot they must have voted straight. So the greater the fraction of straight votes, the fewer remaining opportunities to pick up a split vote.
So I think this analysis is wrong.
However, maybe there's different approach that forms a proxy and looks for unexpected deviations. Of course, maybe not, because if the 'steal' was done right it would apply the same effect equally to straight and split tickets.
I tried a few other plots, but I'm a little tired so I'm not sure I have thought them through completely.
One thing I tried was using the actual overall rate of trump votes (straight + split) as the proxy and then looking at the breakdown between straight and split. This shows another clear linear relationship, as the overall popularity of trump increases, the fraction of straight ticket votes to split ticket increases linearly.
In that case the Y axis is straight_r_count / total_votes - split_r_count / total_votes and the X axis is (straight_r_count+split_r_count) / total_votes
I'm not sure whether or not this is significant. (but i am tired :) . If you look in the github this plot is '/tmp/fig3.png' and the matlab code is in /tmp/plot.m. the input CSV file is in tmp/oakland-mi-subset.csv which selects some of the columns from the 2 sheets.
Whether this is some social effect of independent voters or nevertrumper or got knows what.. or a "steal",,, i don't know.
According to the wapo article conveniently selected for me by Google, there's no evidence of fraud and the gop poll watchers were just complaining racist scaredicats.
Walls are closing in! Trump is a cornered cat. Where do they come up with this shit?
If you want to play with some of this data, i put up a github here
https://github.com/vermiculita/expert-carnival
with the data from oakland county. I tried to replicate shiva's plot and also some other variations
BTW if you want to play with some of this data, i put up a github here
https://github.com/vermiculita/expert-carnival
with some different plots I generated. I'm not yet convinced this linear pattern isn't an artifact of the analysis.
I was curious about this too so I downloaded the source data for Oakland county and put up here:
https://github.com/vermiculita/expert-carnival
if you want to play with the data it's there for one county.
i tried plotting this 2 ways
first the way i think shiva plotted it:
the rate at which straight ticket voters vote R (the measure of 'republicanism') vs the deviation of invidividual ballot voters rate of voting trump from the rate of straight tickets voting R.
then, a slightly different approach - compute the rate of R votes in total (whether straight ticket or individual) as the metric of republicanism, and compare the deviation of the share of straight ticket voters vs the share of indivitual voters.
the results were pretty much the same, but maybe they have the same problem.
i get what you're saying - R voters have to choose between straight ticket and not.. and if more go straight there are fewer remaining to go individual.
but i guess what i'm more interested in is, is there any valid analysis we CAN make here to unpack some kind of manipulation? seems like if the fraud was done right it would have accounted for straight ticket and wouldn't be easily caught out, except by actually manually counting the ballots.
Thanks! Indeed I was able to find for example Oakland County MI which was one of the ones he shows in the video, that led me to here:
https://results.enr.clarityelections.com/MI/Oakland/105840/web.264614/#/summary
which seems to be operated by Scytl
I was able to download an XLS file.
To me the question is whether they can somehow keep scorecard a secret. Even if the perps walk, the truth must be proven because that will lead to a true awakening in which their power is in jeopardy.
I think it will be difficult to hide it without massive numbers of carefully forged ballots.. so I hope they can't manage it. The ballots are distributed all over and it's not just a matter of extra ballots but much more complex I think.. so I hope they can't pull it off.
Also don't forget to throw them all in jail
My guess is this move is more about destroying evidence of cheating and forcing a tainted trump win, than about winning themselves.
Fortunately I imagine the destruction was localized to urban shitholes
PowerPoint make it pretty easy to scale and superimpose images, define a color as transparent, etc. At least the really old versions do.. newer ones I find harder to use
Yeah.. I'll keep working on it. There are some decent free tools but they all take some time and effort to learn. PowerPoint is pretty easy to use for stuff like this
Don't let that prevent you from improving upon it :). Please steal and improve! It needs occupants
Just realized it needs a picture of Joe/Blow in the office
Maybe useful for high risk people. Not much point in anyone else taking it though
I really hope the dominion fraud planning didn't somehow account for this possibility! It would be a big lift to forge the real ballots. If this provides absolute proof of fuckery it will be epic
They should hand count in batches with a machine count to check each batch and recount the batch if different
Even if the counting is electronic, each observer could use their own equipment to verify the count, or spot check it.
Definitely all ballots should be on paper and retained for verification.
Voter should be able to verify their ballot is part of the count but not be able to prove how they voted.
BTW, I did another analysis last night, which had an even stronger linear correlation (I am hoping someone can sanity check this result):
The analysis was:
X axis is (straight_r_count+split_r_count) / total_votes -- the fraction of all votes that are for trump
Y axis is straight_r_count / total_votes - split_r_count / total_votes -- the difference in fractions between straight ticket and split ticket
Here is the resulting plot:
https://github.com/vermiculita/expert-carnival/blob/master/tmp/fig3.png
I can't think of an explanation for this trend, but I'd like to get some other eyes on it to make sure I didn't make a mistake.