170 Registered Voters in Ohio’s 12th District Listed as Over 116 Years Old

[carpedm9587]

Can you point to where you got this? I've been trying to replicate it, but I haven't gotten the number involving the soc security number to match. I've been using the collision probability formula found here https://en.wikipedia.org/wiki/Birthday_problem, which seems sensible. It works for the birthday/year number(setting d to 365*70), but the soc sec number(setting d to 365*70*9999) is really low(on the order of .01% for 223 people), and no other combination (such as 9999 and 70 or 9999 and 365) matches it either. What am I missing? Messing around a bit, I found 36500 for d produces ~50% for 223 people, which seems like an odd number to choose. Is it only the last two digits of the soc security number that are checked? I'm rather confused now.

Wow - nice catch. I didn't get it anywhere - I did the calculations myself, and I made a bad math mistake. I added the 10,000 combinations to the 36500 days instead of multiplying (rookie error). So you are correct, there is a very small chance for a given name at 223 people. You hit 5% at 5,121 people, and 10% at 7,338 people, and 20% at 10,679 people. It accelerates (as you know) until hitting 50% at 18,821 people. So the odds are smaller than cited, but still produce enormous potential outcomes in a population of 325M. It's REALLY hard to find name distribution information, but this site lists the most popular names and has numbers from 26K to 38K for the ten most popular. One can only guess at how many names in the U.S. have more than 10,000 people who have that name, but a quick check of the local phone book can be an indicator. Anyone name that appears more than three times is a decent probability of having more than 10,000 people with that name, especially if the phone book covers and area with more than 100,000 people in it.

[carpedm9587]

It would certainly help your case if you explained exactly how the article in question is unfairly biased, or what information they left out that would change the thrust of the story.

I've already done both, if you reread my original post.

And for your information, I don't blindly trust everything I read on Breitbart, nor are they my sole source of information. I just like referencing them here because it triggers you liberals.

Trigger? Oh I have triggers, MM - no question about it. This is definitely not one of them. But keep on keepin' on, if it tickles you so. I'm pretty sure the Internet term for that philosophy is "trolling."

As to your claim to reading other sources and not blindly trusting Breitbart, that's is a little hard to believe. Brietbart is the primary source you post, and I have never seen you question anything they write. So I don't have much to go on to accept your claim.

[JimL]

Prove it.

You are the proof of your own inanity.

[JimL]

So you are saying that the "majority side" are too lazy or stupid to get valid ID cards? Maybe that is why they vote democrat.

No, but being a conservative/republican I'm not surprised at your comprehension difficulties.

[Pluto]

Wow - nice catch. I didn't get it anywhere - I did the calculations myself, and I made a bad math mistake. I added the 10,000 combinations to the 36500 days instead of multiplying (rookie error). So you are correct, there is a very small chance for a given name at 223 people. You hit 5% at 5,121 people, and 10% at 7,338 people, and 20% at 10,679 people. It accelerates (as you know) until hitting 50% at 18,821 people. So the odds are smaller than cited, but still produce enormous potential outcomes in a population of 325M. It's REALLY hard to find name distribution information, but this site lists the most popular names and has numbers from 26K to 38K for the ten most popular. One can only guess at how many names in the U.S. have more than 10,000 people who have that name, but a quick check of the local phone book can be an indicator. Anyone name that appears more than three times is a decent probability of having more than 10,000 people with that name, especially if the phone book covers and area with more than 100,000 people in it.

Excellent, data is now in agreement. This definitely indicates false positives are likely, and need to be accounted for. To do that, I found https://math.stackexchange.com/quest...-of-collisions which provides a way to calculated the expectation value of the birthday problem. This seems reasonable, though it does count people involved in duplicates rather than duplicates themselves.(So the number needs to be divided by 2 for comparison to Rogue6's number) For reference, this produces ~1.3 people in duplicates for 18821(Which is reasonably close to a naively expected '1'). The most common name(38,313 from your cite) only produces 5.7 people in duplicates, which provides the upper limit of contribution for a particular name. At this point, I think I went a bit overboard.

I found two possible sources for more specific name numbers:https://genealogy.stackexchange.com/...eople-per-name and https://stackoverflow.com/questions/...f-person-names. Unfortunately, while the first talks about how the number decays(resulting in only ~2500 having more than ~3800 duplicate names if the numbers are scaled to the US population), it doesn't actually give the distribution for the whole thing :/. It also suffers from the source, a record of deaths in the past 70 years, which is a bit odd for name extraction. We can say there is probably an upper limit of 12500 people with duplicates(If I take the ludicrous assumption that all 2500 high count names contribute 5 people. 3800 name duplications only produces an expectation of ~.06). It is probably much less.

The second link points to name data from a census record in 1990, where the first and last names are unfortunately unlinked :/. Wanting a bit more a specific number than above, I basically multiplied these percentages in a pairwise fashion and then divided by two(male and female names are split), which produced a distribution of names that is likely highly inaccurate, but at least it's something :/. Finding the expectation of these using the US population and adding them up, I determined we can expect ~820 people with duplicate names. In this schema, the most common name is ~55000 duplicates, so it's a bit high, but I don't know if there are any other name pairs that are abnormally low. As an aside, if I use the number of people who voted in the 2016 presidential election(I think that's the number I'm using(138 mil). If not, it's close, and I'm not redoing my calc) I get around 147 people with duplicates.

The upshot of this is even the more extreme approximations(~6000 duplicate entries, the 12500 divided by 2 for comparison with Rogue6's number) are significantly lower that the amount of double voting cases that Rogue6's source has indicated.(~35000) If we pick a more reasonable approximation and go to total number of votes instead of population, the duplicates might even fall below triple digits. So while they it doesn't look like they did due diligence in checking for the effects of the birthday problem, it doesn't look like it negatively affects the import.

Math is definitely fun .

[JimL]

He's an interesting case study in "repeating the same thing over and over hoping it becomes truererer".

And y'all are interesting cases of either stupidity, brainwashing or both. You wouldn't know that of course, because not knowing is part of the condition.

[Sparko]

No, but being a conservative/republican I'm not surprised at your comprehension difficulties.

The DNC even required photo ID's to attend their convention.



I guess they don't want any low-life poor people who can't afford Photo IDs to attend, huh?

[One Bad Pig]

And y'all are interesting cases of either stupidity, brainwashing or both. You wouldn't know that of course, because not knowing is part of the condition.

CEPG0BtWoAAlovd.jpg

[carpedm9587]

Excellent, data is now in agreement. This definitely indicates false positives are likely, and need to be accounted for. To do that, I found https://math.stackexchange.com/quest...-of-collisions which provides a way to calculated the expectation value of the birthday problem. This seems reasonable, though it does count people involved in duplicates rather than duplicates themselves.(So the number needs to be divided by 2 for comparison to Rogue6's number) For reference, this produces ~1.3 people in duplicates for 18821(Which is reasonably close to a naively expected '1'). The most common name(38,313 from your cite) only produces 5.7 people in duplicates, which provides the upper limit of contribution for a particular name. At this point, I think I went a bit overboard.

I found two possible sources for more specific name numbers:https://genealogy.stackexchange.com/...eople-per-name and https://stackoverflow.com/questions/...f-person-names. Unfortunately, while the first talks about how the number decays(resulting in only ~2500 having more than ~3800 duplicate names if the numbers are scaled to the US population), it doesn't actually give the distribution for the whole thing :/. It also suffers from the source, a record of deaths in the past 70 years, which is a bit odd for name extraction. We can say there is probably an upper limit of 12500 people with duplicates(If I take the ludicrous assumption that all 2500 high count names contribute 5 people. 3800 name duplications only produces an expectation of ~.06). It is probably much less.

The second link points to name data from a census record in 1990, where the first and last names are unfortunately unlinked :/. Wanting a bit more a specific number than above, I basically multiplied these percentages in a pairwise fashion and then divided by two(male and female names are split), which produced a distribution of names that is likely highly inaccurate, but at least it's something :/. Finding the expectation of these using the US population and adding them up, I determined we can expect ~820 people with duplicate names. In this schema, the most common name is ~55000 duplicates, so it's a bit high, but I don't know if there are any other name pairs that are abnormally low. As an aside, if I use the number of people who voted in the 2016 presidential election(I think that's the number I'm using(138 mil). If not, it's close, and I'm not redoing my calc) I get around 147 people with duplicates.

The upshot of this is even the more extreme approximations(~6000 duplicate entries, the 12500 divided by 2 for comparison with Rogue6's number) are significantly lower that the amount of double voting cases that Rogue6's source has indicated.(~35000) If we pick a more reasonable approximation and go to total number of votes instead of population, the duplicates might even fall below triple digits. So while they it doesn't look like they did due diligence in checking for the effects of the birthday problem, it doesn't look like it negatively affects the import.

Math is definitely fun .

Yes, it is - and I appreciate your attempts. I wish I had the time to dig into the math on that site and verify it, and to trace through all of your work above. Alas, I don't. Lacking that time, I have no choice but to put my objection in abeyance until I do. So I am going to accept your math.

That cause me to go looking for this "study" Rogue cited. If it said what he claimed it said, I would have to adjust my position on voter fraud. The fact that it was NC, which was just slapped down for engaging in an attempt to implement VoterID in a way that "targeted minorities with surgical precision," had my suspicions on high alert. I found that what Rogue presented is actually NOT what the study showed. He messed up the numbers, or at least part of his claim. According to this article, only 765 matches were found nationwide using first and last name, date of birth (DOB), and last four digits of the social security number. The 35,750 number he cites were matches nationwide for only first and last name and DOB. When you drop out the Soc, my original numbers are back in play. This is only one of many articles that cites this discrepancy, so it appears to be accurate.

Second, the match works assuming randomness. If there is anything that breaks randomness, it has potential to change the math. On thing that breaks randomness is the exact thing this thread was discussing: states that did not require birth dates prior to a particular time and simply used a default/fill-in (e.g., 1/1/1900). That can potentially significantly spike matches if it occurs with sufficient frequency. Unfortunately, I have no way of showing that it has or has not occurred.

This is why, IMO, the entire "voter fraud" thing is simply a canard. The fact is, showing that voter fraud is occurring to any significant is difficult, and it has been shown to occur only extremely rarely. That is why I would resist any attempt to solve this "problem" (which cannot be shown to exist) with a solution that disenfranchise voters at a greater percentage. There is a simple solution: grandfather existing registered voters, implement a nationwide voterID system, and interlink the information databases of the various government agencies to automate purging of dead people and adjust for voter movement. Indeed, if the databases themselves were standardized, and interlinked with one another, they could self-purge relocated voters without any human intervention.

[Cow Poke]

The DNC even required photo ID's to attend their convention.



I guess they don't want any low-life poor people who can't afford Photo IDs to attend, huh?

Liberals are consistently hypocritical.