From the Gun Guys taken from the VPC press release.
States with the Five HIGHEST Per Capita Gun Death Rates
- Louisiana--Rank: 1; Household Gun Ownership: 45.6 percent; Gun Death Rate: 19.87 per 100,000.
- Mississippi--Rank: 2; Household Gun Ownership: 54.3 percent; Gun Death Rate: 18.32 per 100,000.
- Alaska--Rank: 3; Household Gun Ownership: 60.6 percent; Gun Death Rate: 17.62 per 100,000.
- Alabama--Rank: 4; Household Gun Ownership: 57.2 percent; Gun Death Rate: 17.55 per 100,000.
- Nevada--Rank: 5; Household Gun Ownership: 31.5 percent; Gun Death Rate: 16.21 per 100,000.
States with the Five LOWEST Per Capita Gun Death Rates
- Hawaii--Rank: 50; Household Gun Ownership: 9.7 percent; Gun Death Rate: 2.82 per 100,000.
- Rhode Island--Rank: 49; Household Gun Ownership: 13.3 percent; Gun Death Rate: 3.51 per 100,000.
- Massachusetts--Rank: 48; Household Gun Ownership: 12.8 percent; Gun Death Rate: 3.63 per 100,000.
- Connecticut--Rank: 47; Household Gun Ownership: 16.2 percent; Gun Death Rate: 4.27 per 100,000.
- New York--Rank: 46; Household Gun Ownership: 18.1 percent; Gun Death Rate: 5.07 per 100,000.
What's your opinion? Why is it always so difficult to get the gun owners to admit the that which is so obvious?
Please leave a comment.
Look! Chicago! Daley! Bloomberg!
ReplyDelete--JadeGold
And all the mass shootings in the UK!
ReplyDeleteLaci
Maine has only 8.63 but didn't you just have a post that Maine is out of control and there were loopholes on the lose or something?
ReplyDeleteI wonder what these numbers look like if your remove suicides?
Your numbers are messed up, Mike. North and South Dakota as far as I know haven't had a gun death in several years, yet they aren't on the list.
ReplyDeleteSomething tells me the VPC numbers are wrong.
They also don't cite their sources in the study. Big ethical no-no.
But then again gun control groups are very anti-ethics, hence why so many of their members are in prisons right now....
New Hampshire named safest state in the country
ReplyDeleteApril 5, 2010
Boston Globe
CONCORD, N.H.—New Hampshire has been named the safest state in the nation for the third year in a row.
In a report released Monday, Washington-based CQ Press announced its national crime rankings.
Weerd demonstrates his awesome reading skillz.
ReplyDeleteActually, if he'd looked at the link, he'd find both North and South Dakotas listed (8.94 (34) and 6.54 (42), respectively).
Weerd also claims a big ethical no-no. But had he read he'd see the VPC cites the CDC and where they obtained state firearm ownership data.
Weerd has committed a big literacy no-no.
--JadeGold
More cars, more car deaths.
ReplyDeleteMore pools, more pool deaths.
More snakes, more snake deaths.
More hurricanes, more hurricane deaths
It doesn't take a genius to realize that the more of something you have, the more likely you are to be killed by it.
JadeGold, Thanks for straightening Weer'd out. I hate to do that.
ReplyDeleteFWM, You're right I was denigrating Maine the other day. But what's your point about suicides? Serious question.
Wow. Are middle-school level statistics the best they can do? Even Lott can make up numbers better than the "VPC."
ReplyDeleteMike, you appear to have no background in this subject, so take it from me: these numbers show correlation without any attempt to show causation.
This is the high school level textbook I recommend to people who want to learn basic econometrics.
http://en.wikibooks.org/wiki/Econometric_Theory/Regression_versus_Causation_and_Correlation
Anonymous said, "so take it from me:"
ReplyDeleteYou? Who are you?
Somebody who had to learn this stuff to graduate university, is who.
ReplyDeleteTrusting me is also the alternative to reading this epic snooze of a post, but you asked for it.
Lying with statistics 101:
"He uses statistics as a drunk uses a lamp-post: for support rather than illumination."
Post 1 will explain the problems with this data, and “simple correlation” in general.
Post 2 will explain the advanced methods used by the NRA to make their flawed cases.
Post 3 will attempt to explain the proper methods real statisticians use to demonstrate causation.
1)
Let me start with a simple one. Somebody tells you that, “as a majority of black Americans test into the lowest half of the IQ range, being black obviously causes low IQ.”
Now, most people could not explain why this was wrong, other than by saying “I just don’t believe it.”
Although it is not apparent without using regression analysis, the IQ data set contains virtually no information on a link between race and intelligence.
As most people cannot do this, claims based on “raw” statistics indicating correlation are a favourite tool of propagandists. Most of the “proof” put out by the Brady folks and low level gunners comes in this form. The NRA has better-educated liars, so they use different methods (see post 2)
As you no doubt learned in high school, simple linear regression analysis determines the level of correlation between a dependent (IQ) and an independent variable (race). But if we believe that IQ can be affected by other things, such as education, poverty, discrimination, etc. we have a problem: being black is also correlated with the other independent variables, so we have less unique information than we thought we did: does being black make you dumb, or does being poor and marginalized make you dumb?
The infamous book “The Bell Curve” uses this method to “prove” this exact claim. The author simply removed education from his list of independent variables to get what he wanted to see. All the correlation between education and IQ (huge!) got shifted into the correlation between being black and IQ--as if being poorly educated was _entirely_ due to skin colour.
Make a mistake like this in intro econometrics, you’d be lucky not fail the class. Do it in a book, and you make a lot of money from racists.
This is as best as I can explain without using un-typeable math, but you can look it up in any statistics textbook.
So, any time somebody gives you data and says “the conclusion is obvious,” they are lying to you, or were misled themselves. Presenting data in this format is the same as the regression error, but it is harder to spot the critical mistake.
The “information” provided by the VPC is a classic example.
This is not to say that they do not have a case! They may be entirely correct, but their methods cannot prove it.
2)
ReplyDeleteNow, because the raw data doesn’t make their case, NRA backers have to use sneakier methods. They are numerous and complicated, so I will stick to the best and evilest here.
John Lott is a perfect example. In his book “More Guns, Less Crime,” he attempts to prove that gun control increases the rate of violent crime. He does this by making gun control legislation (carefully defined to maximize his results) an instrumental variable in a time series: cities with gun control appear to have more violent crime.
The trick is ignoring that cities with high violent crime are more likely to pass gun control legislation. Untangling autocorrelation is really difficult, and he does not attempt is, as his h-statistics would show if he bothered to include them in the appendix!
Basic regression analysis assumes lack of heteroskedasticity, among other strict conditions. Lott ignores this. You can make a case for it being unimportant, but he fails to do this, or even acknowledge the issue. Since the 80’s folks have developed new techniques for dealing with the problem, but the methods are too complicated for a piece of agitprop like Lott’s book.
He also pulls the stunt explained in post 1, but hides it better—in ways it gave me a headache trying to untangle.
Lott’s methods are more insidious than the Bradys’ because it appears to be science, and thus beyond criticism. There are many ways to torture statistics to get whatever result you want, and if your work is peer-reviewed you will be lucky to get ripped apart for it. Why lucky? Because if it slips through everyone and his mother will publish papers politely and condescendingly crushing your study, and suggesting that you are either past your prime or need to go back to school.
People with an agenda to push usually publish books instead, so they can rely on a dumb audience who already agrees with them.
This is why you should never trust statistical analysis you see in a mass-market publication, especially if you know the author has an agenda.
The major warning signs of this kind of thing are claims of extraordinary significance (“cell phone use explains 200% of cancer cases!”) discovered with simple experiments (“cold fusion in your backyard!”), coupled with conspiracy theories to explain why you or other researchers have not published in journals.
Lott was embarrassing enough that both pro and anti gun-control economists got together to disprove his theories. Of course, they didn’t get much press outside of academia.
3)
ReplyDeleteSo, how do real statisticians handle the issue?
The answer is that they usually don’t, for a number of reasons. It is very complicated, hard to get firm conclusions, and too easy to be misinterpreted by the press (which can ruin your reputation).
Some people have made a stab at it, though, because difficult problems always attract new ways of solving them. If you have a new method (that gives different results), but don’t want to step on anyone’s toes by contradicting work they have already done, you can do an original study as a kind of proof-of-concept.
One really neat trick is the “natural experiment” technique. Unfortunately, statisticians are not allowed to lock a few thousand people two buildings—one full of guns, one full of butter—just to see what happens. What we can do, however, is look for situations where the independent variable in question was somehow changed with everything else held constant. This is how “real” experiments work: to test the volume of a gas you must keep the temperature constant while increasing pressure, and so on.
We’ve already seen that what laws get passed is at least partially determined by the current situation, which makes this difficult.
Judicial review, however, is a wonderful opportunity. Imagine that two cities with known characteristics have the same gun laws, and that the laws are appealed to different circuit courts. If the 3rd court strikes down the laws, and the 9th court upholds them, we now have gun laws on the books that have nothing to do with the actual situation, because for statistical purposes judges’ decisions are essentially random. :D
Now we can even use Lott’s method without error, because we can control for all other characteristics of the cities!
Unfortunately, the Supreme Court quickly steps in to resolve disputes between circuit courts. Since the various effects of gun control probably take years to materialize, we can only get short term results this way, but the principle is sound.
There are many situations where the natural experiments can help, and it’s not the only tool.
Alright, if you made it through these posts you now have a basic understanding of the power and limitations of the analysis of correlation. You know why the data you presented above proves nothing. Best of all, you now know what you need to make really good arguments in favour of gun control.
Or you could just keep posting nonsense from the Brady Bunch, and your hecklers will keep posting nonsense from Lott and the NRA, while the rest of us give up and ignore you… That works too.
I just want to try to give you some idea of what you’re missing by only reading “news” and analysis that you either agree with or get furious at.
Apologies for two typos:
ReplyDeleteIn post 2 "instrumental variable" should read "causal variable."
His use of instrumental variables is something else entirely...
The second paragraph of post 3 should read "thousand people into two buildings"
Mr. Anonymous, No problem with the typos. Whoever you are, you've got that verbosity thing going on, man. I'm a little busy these days, maybe next Saturday, oh, no, I just remembered, I have to do something. Ok, I'll tell you what, over my summer vacation I'll read your comments and get back to you.
ReplyDeleteI did check one thing, heteroskedasticity really is a word. I guess that's what I get for dropping out of high school, huh?
you should never trust statistical analysis you see in a mass-market publication, especially if you know the author has an agenda.
ReplyDeleteWell that throws every single anti-gun "study" ever done out the window.
Thanks for taking the time to read my looongpost, Other Mike.
ReplyDeleteUnfortunately, knowing who it is safe to cite gets harder and harder as debate on a subject gets polarized.