|
*** UPDATED x1 *** Bending the laws of statistics
Thursday, Jun 19, 2014 - Posted by Rich Miller * All emphasis added. From a June 5, 2014 Chicago Tribune editorial…
* From a June 17, 2014 Tribune editorial, “Illinois elections board is rushing to kill a people power amendment”…
* From a June 19, 2014 Tribune editorial…
* As I explained to subscribers yesterday, this is clearly not a “red flag.” And any sane, rational person without a not so hidden agenda would easily understand this. If you click here and download the spreadsheet provided to media outlets, including the Tribune, you’ll see a very easy explanation for those two examples that the Tribune keeps harping on. Examiners looked at 25,000 petition entries, selected at random via computer. They struck 13,807 as invalid, for a failure rate of 55 percent. * If you can’t download the spreadsheet, here are the numbers. The first column is the examiner number, the second is the total number of petition entries examined by each examiner, the third column is the total valid petition entries by that examiner and the fourth is total invalid entries found by each examiner…
Toss out Examiner 38 because s/he only looked at one petition entry. You get an average of 676 entries examined and a median of 711. * Now, onto the Trib’s goofiness. The Board of Elections staffer who “disqualified only 17 percent” (or 16 percent, depending on which day you read the Trib) examined just 92 petition entries. The staffer who “disqualified 86 percent” looked at just 183 entries. Both of those people looked at far, far less petition entries than the average or median. Statistics seems like a difficult topic, but let’s put it in a political way to make it easy to understand in our world. If you have 37 people making phone calls for a poll, would you publish only the results from two workers who made 0.7 percent and 0.4 percent of the total phone calls? Of course you wouldn’t. What matters is the overall number, not cherry-picked results from a couple of people who looked at a tiny amount of petition entries. Those two examiners darkly singled out by the Tribune were the most serious deviations, but deviations are naturally expected among all examiners, particularly considering they weren’t just matching up signatures, but also looking to see if people were registered to vote (subscribers know more about that and I’ll get into it in due time here). Different randomly selected stacks are just gonna have different results. Plain and simple. Not all circulators were equal. Not all regions covered were equally managed (and some regions were checked in bunches on certain days). Not all signers were honest. You’re just naturally gonna see differences when looking at any individual samples, and particularly so when you single out some of the tiniest samples. And, heck, even if there is a real problem with those two, toss them out and the remap folks are still a very, very long way from making it onto the ballot. * The bottom line here is that the Tribune, out of either ignorance or malice, has repeatedly used an easily disprovable statistic to claim that the Illinois State Board of Elections has corruptly stacked the deck against the remap reformers. It needs to stop. *** UPDATE *** OneMan in comments…
|
 |
|
|
|
|
- steve schnorf - Thursday, Jun 19, 14 @ 12:08 pm:
I would say “Oh how the mighty fall”, but the Trib has long since lost its credibility as “mighty”
- Willie Stark - Thursday, Jun 19, 14 @ 12:08 pm:
Your work really demolishes the Trib’s pretensions that it is an ethical guardian of the broad public interest. It also highlights the degree to which it has embraced the role of political actor and handmaiden to the Rauner candidacy. The snarky, hyperbolic and hysterical tone of today’s malicious editorial, and it’s malice, not ignorance, takes one’s breath away. A great city does not have a daily newspaper equal to its greatness.
- A guy... - Thursday, Jun 19, 14 @ 12:11 pm:
Refer to Mark Twain’s quote about statistics. Simple and enduring.
- Skeptic - Thursday, Jun 19, 14 @ 12:14 pm:
Who are they to let a few facts get in the way of a good byline?
- A guy... - Thursday, Jun 19, 14 @ 12:14 pm:
===A great city does not have a daily newspaper equal to its greatness.===
Fewer and fewer “great” cities do. On the other hand the city and newspaper could hardly claim greatness these days. They aren’t on vastly different planes to a lot of people.
- Oswego Willy - Thursday, Jun 19, 14 @ 12:19 pm:
To the Post,
===* The bottom line here is that the Tribune, out of either ignorance or malice, has repeatedly used an easily disprovable statistic to claim that the Illinois State Board of Elections has corruptly stacked the deck against the remap reformers. It needs to stop.===
Even with editorial license, you can’t “make” numbers be the argument and then not fully comprehend what the numbers say.
It is a major fail of making a point, when you have no point in your argument.
- OldSmoky2 - Thursday, Jun 19, 14 @ 12:21 pm:
In other words, half of the daily newspapers in Chicago skew the numbers to mislead their readers.
- Keyser Soze - Thursday, Jun 19, 14 @ 12:22 pm:
Apart from signatures not belonging to a registered voter, what other criteria were used by the examiners to reach their conclusions? I’m a math guy, so this is fascinating stuff.
- Rich Miller - Thursday, Jun 19, 14 @ 12:24 pm:
===what other criteria===
Subscribe.
- lake county democrat - Thursday, Jun 19, 14 @ 12:26 pm:
Rich, there still seems to be a great deal of unexplainable statistical variation here. Fine, throw out those two examiners who were light (for whatever reason) on the number of petitions they examined. Compare Examiner 3 with Examiner 8 - one has an approval rate of 56%, the other 36%. With each doing approx 1000 ballots, that can’t be because one random stack just happened to be higher quality than the other - the chances of that with that sample size is tiny. That’s different standards of judging. Or if we take the “middle” examiner (#19) the approval rate is over 50%. In fact, there’s an objective way to judge this: the statistical variation of the samplers.
Also, you don’t HAVE to believe the Board corruptly stacked the deck to believe that the Board is wrong here - it’s possible that they gave insufficient guidance and supervision to the examiners. Again, has the press had a chance to evaluate the “low” examiners? Aren’t these ballots public records? It would be helpful to see what the Yes Maps people were challenging as unfair rejections.
- Mighty M. Mouse - Thursday, Jun 19, 14 @ 12:27 pm:
===The bottom line here is that the Tribune, out of either ignorance or malice===
Please. They’re not THAT ignorant. It’s malice.
- Rich Miller - Thursday, Jun 19, 14 @ 12:30 pm:
LCD, as subscribers already know, signature matchups and legibility represented a tiny, tiny fraction of the petition entries tossed by the Board staff. The vast majority were tossed for purely objective reasons. And there was a significant petition quality discrepancy between Downstate and the Chicago area and many of those were examined separately, explaining other deviations.
- R.T. - Thursday, Jun 19, 14 @ 12:38 pm:
If anyone at the Trib ever ciculated petitions or participated in a petition challenge, they would know there is nothing unusual about the desparities they are freaking out over in repeated editorials. The quality of signatures is dependent on the work of the circulator and that quality often varies widely. If one circulator stands on the L platform at the DePaul U. stop, he or she is going to get a lot of bad signatures because of the transient nature of the people — both student and nonstudent — who live in that neighborhood. I would expect that 80 percent of those L platform signatures would be from people who are not registered to vote.
But if a circulator took the time to walk door-to-door with a poll sheet listing registered voters in the 41st Ward — an area made up of single-family homes where a ton of city workers and retirees live — then well over 80 percent of the signatures would be valid.
It’s nothing sinister, just Politics 101.
- Tom Joad - Thursday, Jun 19, 14 @ 12:38 pm:
Good work, Rich.
- OneMan - Thursday, Jun 19, 14 @ 12:41 pm:
Doing some quick math in the spreadsheet and eliminating any examiner above #30 due to their small sample size, one sort of interesting thing comes up…
I show a average failure rate of 53.38% roughly with a standard deviation of 9.5% so using a normal distribution.
+2 Stddevs is 72.24% -2 stddevs is 34%
We have one examiner who fails below -2STDEV but none that are +2 Over…
So all in all it seems to pass statistical muster, a touch over half are over the average failure rate.
If a was the pro folks I would take a look at Examiner 12 a touch and see if I could get examiner 30 to be cloned, but all in, it is statistically clean.
- The Doc - Thursday, Jun 19, 14 @ 12:49 pm:
Three editorials about the remap, zero about Rauner’s misleading and vacuous budget platform.
Dold, McQueary and the rest of the lot deserve all of the ridicule and then some more.
Shameful.
- Ray del Camino - Thursday, Jun 19, 14 @ 12:51 pm:
I know the goo-goos have been told numerous times to hire pros. They won’t listen. It’s the only way they’re going to get anything done in this state. This will be very disappointing if the remap question doesn’t get on the ballot.
- wordslinger - Thursday, Jun 19, 14 @ 12:56 pm:
If they weren’t a bunch of willfully ignorant hacks, the Tribbies would present evidence before they accuse the people at the SBOE of being corrupt.
Perhaps they could explain as well why the SBOE is in the tank on the map amendment, but not Rauner’s term-limit amendment, which passed muster like stuff through a goose.
Doesn’t make a whole lot of sense, if the process is corrupt.
Could it be the Rauner crew actually was competent and did the job right, while the map people themselves stated that only “above 60%” of their sigs were kosher?
Maybe the map people could explain what they did with “above $2.4 million” in contributions (according to SBOE filings) to turn in such shoddy work.
Some pretty sweet paydays in those filings. But they were goo-goo-sweet paydays.
- A Republican - Thursday, Jun 19, 14 @ 1:00 pm:
The reason the proposal falls short of the necessary numbers is Demoquacks don’t want the map re-drawn. End of story.
- LincolnLounger - Thursday, Jun 19, 14 @ 1:03 pm:
The UIS should do a case study and compare the two initiatives when this is all over. Management, funding, etc. Also, I think we should be talking about the names of the signature-gathering firms involved.
- wordslinger - Thursday, Jun 19, 14 @ 1:05 pm:
–The reason the proposal falls short of the necessary numbers is Demoquacks don’t want the map re-drawn. End of story.–
The map is re-drawn every ten years, Einstein. That quirky Constitution.
- Keyser Soze - Thursday, Jun 19, 14 @ 1:11 pm:
I haven’t given this too much thought but I wonder if the normal distribution comes into play in this one. Questions involving a coin toss, pass/fail or other yes/no responses tend to follow a binomial distribution. A better experiment would be to use an Analyis of Variance (ANOVA) to compare the significance of the reasons for rejection/acceptance among examiners as a test of bias (in the mathematical sense). This is fun stuff for techie-nerds.
- Mason born - Thursday, Jun 19, 14 @ 1:14 pm:
Hopefully the remap folks will take a breath figure out what went so gawd awful wrong and try again. Lord knows they should be able to find plenty of mistakes to correct.
- Wensicia - Thursday, Jun 19, 14 @ 1:15 pm:
The Tribune editorial board is critiquing the examiners as if they were Olympic ice skating judges. Cheap shots with malice instead of factual reporting.
- Snucka - Thursday, Jun 19, 14 @ 1:16 pm:
More fun with numbers:
The remap group can only have a 41% “invalid” rate if they want to stay on the ballot. An astounding 87% of the examiners found a rate higher than that.
Also, why is the board corrupt on this issue but not corrupt when they allow the term limits amendment to survive?
- Snucka - Thursday, Jun 19, 14 @ 1:18 pm:
I guess to answer my (and wordslinger’s) question, one could argue that the term limits amendment is more likely to get bounced by a judge. That’s still not evidence of corruption, but I imagine someone will make that argument.
- Anonymous - Thursday, Jun 19, 14 @ 1:21 pm:
What RT said. If the proponents were testing properly along the way, they should have noticed the failure rate by location early on, and adjusted accordingly. The goal is quality, not quantity. Or shoulda been
- thinkaboutit - Thursday, Jun 19, 14 @ 1:30 pm:
==If a was the pro folks I would take a look at Examiner 12 a touch and see if I could get examiner 30 to be cloned, but all in, it is statistically clean. ==
The stats don’t really matter because you can’t compare apples and oranges. Each page of the petition is different. Each examiner is looking at different pages. It is very possible that one examiner was handed a stack filled with false names or addresses, and the other got sheets that were nearly perfect. One may have a sheet that was signed door-to-door using walk sheets, a very reliable method to find registered voters. The other could have been signed by anyone walking in to a Jewel, a very unreliable method to find registered voters.
- lake county democrat - Thursday, Jun 19, 14 @ 1:31 pm:
Rich - thanks for the explanation.
- Josh Collins - Thursday, Jun 19, 14 @ 1:33 pm:
I think one of the things being lost in all the rhetoric is with all our technology we have a very poor and outdated system for signature gathering and voter initiatives. Why are we only sampling 25,000 signatures when the number needed is 298,000. I can see doing things this way in the 20th century, but here in the 21st century things should be a lot easier to determine voter initiatives.
I also believe that every vote counts and that every signature should count. We don’t determine elections based upon random samples and we should not determine voter initiatives on them either, both are too important.
- Rich Miller - Thursday, Jun 19, 14 @ 1:39 pm:
===when the number needed is===
Yes, let’s move the bar yet again.
- Precinct Captain - Thursday, Jun 19, 14 @ 1:58 pm:
==We don’t determine elections based upon random samples==
Actually, we do have recounts based on random samples of ballots to see if they recorded results accurately. So yes, an election can be determined because of a random sample. It’s not an A (random sample) to B (election result) process, but it’s still use of random sample to determine the eventual outcome, A (random sample) to B (recount) to C (result).
- Demoralized - Thursday, Jun 19, 14 @ 1:58 pm:
None of this would be an issue if this group was incompetent to begin with. This is entirely their fault. Those of you who think this is some big conspiracy and that the election board is a bunch of hacks are complete dopes, with the Tribune being the biggest dope of them all.
- Demoralized - Thursday, Jun 19, 14 @ 1:58 pm:
“wasn’t” incompetent.
- walker - Thursday, Jun 19, 14 @ 2:11 pm:
@A Republican: Try to avoid victimhood and laziness by seeing this review as primarily partisan.
Are you saying the “Demoquacks” wanted the Term-limits and Reform petition from Rauner’s group to pass, and that’s why the Board found it perfectly fine?
Again, don’t blame the refs if you stumbled out of bounds.
For the group to turn in petitions, which even they said they assumed were 40% wrong, is ridiculously risky. Most petition submitters go thru a process of checking and cleaning up petitions prior to submission. The obvious lack of quality here destroys credibility going in.
I have not seen these, but with bad quality petitions it’s never just illegible or hard to identify signatures — it’s more often persons not registered to vote at all, persons not resident of Illinois, duplicates, same signature for multiple different names, pages of obvious creations of non-persons by your paid petition passers, cartoon character or obscene signatures, blank or incomplete lines counted as if they were filled out, and on and on. You can roughly eyeball these things and know you’ve got a big problem with some of your geographies or petition passers. Different reviewers, looking at different sheets, will legitimately always have different quality outcomes.
BTW, I support remapping and view this as a disgrace.
- Bill White - Thursday, Jun 19, 14 @ 2:20 pm:
I agree with walker
I do support independent maps - although Illinois is far less abusive than many GOP states, however, to trash the refs because your team fumbled the ball is seriously not cool.
- in the know - Thursday, Jun 19, 14 @ 2:21 pm:
-A Republican-
I too am a Republican and I signed the petition. But there is no way to fairly and accurately politicize this mess. The number of signatures from unregistered voters and from those who were not registered at the address listed is appalling. And as a part of a ballot integrity team for the GOP, we would object to these folks voting (other than a provisional ballot in the case of the wrong address). If the “Demoquacks” had done this we’d be rightfully screaming to high heaven to block their votes OR their petition signatures. So be principled enough to admit when our side messes up and quit embarrassing yourself and our party with your mindless partisan drivel. Its hard enough to be a GOP members sometimes without your brand of troglodyte commentary.
- A Republican - Thursday, Jun 19, 14 @ 2:48 pm:
@in the know
Thank you for the cautionary warning about embarrassing myself.
Long ago, I learned it is still a free country. So I shall say whatever I wish, thank you.
- Demoralized - Thursday, Jun 19, 14 @ 3:05 pm:
@A Republican:
You’re right. You are perfectly free to say whatever wish - dumb or not. Some of us tend to ignore dopes when they start using terms such as “Demoquaks.”
Also, you might want to know this little tidbit - the State Board of Elections is evenly divided between Republicans and Democrats.
- Matt - Thursday, Jun 19, 14 @ 4:00 pm:
Adding to what R.T. wrote above, I helped a friend with a petition challenge last fall. I spent half a day at the county clerks office checking signatures submitted by his opponent. Some of the petition sheets were rock solid, all but one or two of the signatures were valid. But some sheets were equally bad, with only two or three valid signatures.