Tuesday, August 27, 2013
Dom Raso on Background Checks
In the nonsense department, this guys is worse than Colion Noir, and that's saying something. He said 76,000 failed the background check in 2010 but only 13 of them "were legally blocked from buying a firearm." In true Colion-Noir form, he tried to confuse the argument by conflating the "blocked from buying a firearm" idea with "actually prosecuted and convicted of committing a felony in order to try and buy a firearm."
The fact is ALL 76,000 were blocked from buying a firearm. This is proof that the background check system, at least where it's required, works extremely well. The question of arresting and charging people for the paper crime of lying on a government form is a different issue and has little to do with the background check system's efficacy. Gun-rights fanatics use this to obfuscate the fact that background checks work.
I should apologize right here to Colion. Compared to poor Dom, Mr. Noir as hip, slick and cool. Dom Raso is a bumbling and embarrassing inferior imitation.
What's your opinion? Please leave a comment.
Subscribe to:
Post Comments (Atom)
No, you missed his point. What a surprise. He's saying that unless the government gives a whole lot more time and money to enforcement, the background check system will be this clunky bureaucracy that gets a lot wrong.
ReplyDeleteWhat we don't know is how many of those 76,000 were names that had been entered incorrectly or were names like someone else's or were some other glitch in the computer.
But you're happy, since those people were denied a firearm. Who cares whether their rights were violated? Who cares if they were actual criminals who got away with a crime? You just want more layers of government to slow things down like some massive denial-of-service attack on our rights.
I missed the point? Listen to the video, Greg. I QUOTED the man.
DeleteYou quoted a piece of the argument, but left out the point.
DeleteI quoted the lying part and I commented on his point which was a sloppy attempt to deflect the attention away from the fact that tens of thousands of people are refused firearms each year due to the background check system.
DeleteAnd of that number, only a handful are prosecuted. We're left to wonder how many were false positives. Of course, no one is checking.
DeleteI have a megapost answer for this exchange we had. This is important stuff, and the thread we were discussing this on is getting pushed pretty deep, so I am bringing it up top:
ReplyDeleteMikeB: “If it weren't for guns, we would have far fewer murders. Do you deny that? This is a yes or no question.”
TS: “Not just "far fewer", there wouldn't be ANY fewer. Do I need to redirect you to my posts on the subject, or do you want to deny the math of correlation coefficients?
MikeB: “Yes, TS, I deny that.”
http://mikeb302000.blogspot.com/2013/08/us-gun-culture-is-corrupting-world.html
Ok, Mike, here we go. This is another attempt by me to settle the issue of how guns relate to murder rates, and the numerous posts I have done concluding that there is zero correlation between gun laws (or gun ownership) when compared to FBI UCR statistics on murders. For reference, here is a link to some of our threads on the topic earlier in the year:
http://mikeb302000.blogspot.com/2013/03/an-argument-for-gun-registration.html#comment-form
http://mikeb302000.blogspot.com/2013/03/states-with-weak-gun-control-laws-have.html#comment-form
And here is a link that explains how correlation coefficients are calculated:
http://en.wikipedia.org/wiki/Pearson_correlation_coefficient
I think part of the problem is I lose you when I start talking about math beyond calculating averages, and that is perfectly fine. Not everyone has a math or statistics background, and you don’t need one to engage in intelligent debate either. But at the same time, you can’t just deny the math, or the data, without making an attempt to refute it. So I have been thinking of ways to at least show you what I did and why, and when you understand it you can properly critique it if you feel something is wrong, rather than just calling my math “pro-gun biased” as you have in the past.
For our case we have 50 states, each with a Brady score, and a murder rate. When you last tried to prove how effective gun control is, you took ten states from the bottom of the Brady score list, and ten states from the top, and calculated the average murder rate. You got a slight (i.e. negligible) difference in your favor and concluded that gun control saved lives. I pointed out the severe flaws in this method- first that you are leaving out 60% of the data, and second, that you didn’t account for what the Brady score is. You essentially assigned them a discrete value of “gun control” vs. “no gun control” even though the Brady scores are not even close to a linear distribution (CA has over 50 more Brady points than the #10 ranked state, and you excluded states with a similar, or even identical Brady score on the “no gun control” end). If you included all states and just split it down the middle it wouldn’t make any sense because most states have low Brady scores, and again it wouldn’t account for what the Brady score actually is. This is why correlation can’t be determined with a simple average calculation, instead there are specific formulas used.
On a basic level, correlation coefficients, such as Pearson, are used to determine the relationship between two or more sets of data. Pearson (r) ranges from -1 to 1. 1 is a perfect positive correlation, and -1 is a perfect negative correlation (when one number goes up, the other goes down). Zero means no correlation, though that doesn’t mean they are unrelated (see the example in the wiki page of scatter plots that look like an “X” or a ring).
(cont.)
No, you lose me in the excessive prolixity of your argument. It's not the numbers it's the words.
DeleteIf you can't make your point more quickly and concisely, you're attempting the old "baffle 'em with bullshit" action, which, I'm sorry does not convince me.
He's showing you the mathematics that support the point he's made repeatedly--namely that there is no correlation between gun laws and rates of murder.
Delete"Too many notes", says the King.
DeleteLook, Mike, I've made my point with the brevity you require a hundred times:
Murder rates are not linked with gun laws.
You say, "I don't believe you". So I show you the numbers, you say, "I don't believe you". I show you the calculations, you say, "I don't believe you". I walk you through step by step how to do the calculations yourself, and now I'm "baffling you with bullshit". Mike, you don't WANT to believe it. You're like a 21st century round earth denier. I can't seem to convince you that 2+2 = 4.
But if you say it's not so, what led you to that conclusion? It wasn't math. If it was, show me how you got there, or give us a link to how someone else got there. Murder rates to gun laws, all 50 states. Can you show me anything that contradicts the results I showed you? I promise I'll address whatever you present. I won't ignore it because it's too verbose, or dismiss it as "baffling with bullshit" without showing you WHY it's bullshit.
Ah! The classic excuse of minds that refuse to learn to understand complex issues.
DeleteTS, you conveniently forget that I responded to you once upon a time with the same kind of detail you use but from different sources. And, do you remember, it supported my argument.
DeleteAt that point, I said "basta." That means "that's enough."
"I forget"??? I addressed your analysis right here in my megapost- the very same comment you are responding to. And I addressed it at the time you made it. Do you see the difference between you and me? I offered a multi-point critique of your methods. You have never even attempted to dispute what I actually did- instead you did something else and called it an answer.
DeleteNotice how I had you start with Louisiana and California in these steps? That was to prove a point, and further rebut what you did earlier this year. Louisiana is a bad data point for me (the worst by a large margin, in fact). So when you compare Louisiana to some other state, and ignore 48 others, it looks good for gun control. If you throw a few more states in, it doesn't quite look so good, but you most likely still have an advantage. That's what you did. You only looked at 20 out of 50 states, and by then the margin in murder rate was down to 0.25 per 100k. By the time you look at all states with a comprehensive analysis, Louisiana is completely cancelled out.
Again, your analysis completely relies on Louisiana. As I pointed out last time, if Louisiana had two measly extra Brady points, they would be out of the bottom ten, and your math would give my side a 0.25/100k advantage. That kind of perilous teetering is indicative of poor methods. Do you know what happens to my answer if you give one state two more Brady points? The r coefficient changes by 0.007.
And you didn't use different sources, you used different methods.
DeleteAgain I'll state that my sources are the Brady Campaign and the FBI.
Remember, what I am asking for is data that links murder rates to gun laws for ALL 50 states, not just 20.
DeleteSince you refuse to see reality, Mikeb, it's clearly not enough.
DeleteAnd if you stick an rd on the end, it's you
DeleteAnother thing you can do is just look at the table of raw 2009 figures I just posted. The Pearson calculation is a comprehensive mathematical comparison of all the states, but you can simply look and see that the numbers are all over the place. I’m sure you’re immediately drawn to the numbers that are on your side as being “proof”, like Louisiana and Hawaii (for opposite reasons). But you can also see that states like New Hampshire and Maryland work against your hypothesis. So let’s throw them into buckets. Which States help you, which States hurt you, and which are neutral? You can make whatever divides you want, the point of the exercise is to visualize why the correlation ends up at zero?
DeleteHere is a list of States with low gun control (under 25) and high murder rates (over 6):
AL 16 6.9
AR 4 6.2
LA 2 11.8
MO 4 6.4
MS 6 6.4
NM 4 8.7
OK 2 6.2
SC 10 6.3
TN 8 7.3
Add to those the list of States with high gun control (over 25) and low murder rates (under 4):
CT 58 3
HA 50 1.7
MA 65 2.6
NJ 72 3.7
RI 44 2.9
NY 62 4
That is a total of 15 states on your side. Now let’s look at how many states have low gun control and low murder rates:
AK 0 3.1
CO 15 3.5
IA 7 1.1
ID 2 1.4
ME 9 2
MN 14 1.4
MT 2 2.9
ND 4 1.5
NE 8 2.2
NH 6 0.8
OR 17 2.2
SD 4 2.6
UT 0 1.3
VT 6 1.1
WA 17 2.7
WI 8 2.5
WY 8 2.4
And states with high gun control and high murder rates:
IL 35 6
MD 45 7.7
MI 25 6.3
That’s a total of 20 states on my side. The remaining states are the ‘tweeners’ with murder rates close to the national average:
AZ 0 5.4
CA 80 5.3
DE 15 4.6
FL 5 5.5
GA 8 5.8
IN 4 4.8
KS 4 4.2
KY 2 4.1
NC 16 5.3
NV 5 5.9
OH 9 4.5
PA 26 5.2
TX 6 5.4
VA 16 4.4
WV 4 4.6
This isn’t a proper way to analyze the data, but it does provide visualization for how the correlation coefficient calculates out to zero. You can see that there are many states that don’t fit the narrative you so desperately want to believe.
PART2
ReplyDeleteNow here is where we are going to have some fun. I trust you have Excel on your computer? I can easily show you what Pearson does with some step by steps and we can finally put this issue to bed.
1) Open a new Excel workbook and type this into cell A1: =PEARSON(B:B,C:C)
Don’t forget the equals sign at the beginning, but you don’t need it when typing in the values in the steps below. This will calculate the correlation between all data points in column B with the respective row in column C
2) Type these values in: B1=1, B2=2, C1=1, C2=2. Cell A1 should be equal to 1, showing that columns B and C have a perfect correlation. If you reverse column C (or B) so that C1=2 and C2=1, you will get an answer of -1 showing they have a perfect inverse correlation. Note, when only using two sets, we can only have an answer of 1,-1, or 0 (which will crash the formula because it will try and divide by zero). If you make C1=100 and C2=100000 or 101, it will still be a perfect correlation. Change C2=99 and the answer is -1. A perfect correlation means all the data goes in the same direction, even if the changes are very small. This is important to realize, because even if gun control had a very small effect, it would still show up with a strong correlation as long as murder rates were consistently changing with gun laws.
3) Now let’s add a third point so that we can see answers that fall in between perfect correlations: B1=1, B2=2, B3=3, C1=1, C2=2, C3=3. This is still a perfect correlation of 1, but make C3=2 and we start getting less than perfect (0.866). Now how much the numbers change relative to each other matter. Make C3=4 and it is still less than perfect (0.982) because C2 didn’t move up as much as C3. Setting C3=1 takes us down to zero, and you can clearly see that there is no relationship between columns B and C, right, Mike? That is an easy to see an example of three data sets that are not correlated at all. Play around with 3 or more sets like this. Punch in numbers so you can see how the coefficient responds. Make the answer go negative. This is a good way to visualize what this calculation actually does, and how Karl Pearson didn’t write “pro-gun bias” into his equation in the late 19th century.
4) Next, let’s put some real numbers in there. Column B will be the Brady score of a given state, and Column C will be the murder rate. Let’s start with your favorite whipping boy State, Louisiana in row 1, and compare it to the Brady Campaign golden child, California in row 2. B1=2 (LA Brady score), C1=11.2 (LA murder rate), B2=81 (CA Brady score), C2=4.8 (CA murder rate). Erase rows 3 and higher. The answer of course, is -1. A perfect inverse correlation that “proves” gun control works (Brady score goes up- murder goes down)…. Or so you’ve claimed numerous times, because you guys love to compare Louisiana to some other state and stop right there as if the 48 other states don’t exist (or you use “gun deaths” as a metric instead of murder rate). But we’re going to keep going, aren’t we?
(cont.)
PART3 (the exciting conclusion)
ReplyDelete5) Add the other 48 states down to B50 and C50 by mining data from FBI UCR and the Brady Campaign website. It doesn’t matter the order, just make sure each row represents a state. What do you get? That is what I did, and showed you the result in my previous posts. As you can see, it is not that complicated, and when you punch in test numbers looking at just three rows, you can easily see what this coefficient represents. If you don’t want to spend the time mining data, I can drop the numbers I have into a post and you can spot check it for mistakes (note: I have 2009 data, and I know 2010 or maybe even 2011 is available now).
6) One last bonus step for you. Copy off your data somewhere and replace B1 with =rand(), making sure to type the equals sign. This is Excel’s random number generator. Copy B1 and paste it into B2 through B1000, and then again for C1 through C1000. This tests the formula for data that literally have nothing to do with each other. They are random. What answer does it give you? Is it a lot like the answer for how well gun control works? Hit F9 a couple of times as well. Each time, it will recalculate with a new set of random numbers. As long as you use enough rows, you answer is pretty much zero every time. Note: murder rates have a tighter group than full 0-100% random number generator, so using only 50 rows in a random generator will get answers even above 0.30 r, or below -0.30 r. But if you recalculate enough times, you’ll see the answer bounces around zero. That’s why I had you copy the random numbers all the way down to row 1000.
Here is the complete 2009 data. Plug it into that excel sheet. Tell me what I did wrong or stop denying the math and data.
ReplyDeleteAK 0 3.1
AL 16 6.9
AR 4 6.2
AZ 0 5.4
CA 80 5.3
CO 15 3.5
CT 58 3
DE 15 4.6
FL 5 5.5
GA 8 5.8
HA 50 1.7
IA 7 1.1
ID 2 1.4
IL 35 6
IN 4 4.8
KS 4 4.2
KY 2 4.1
LA 2 11.8
MA 65 2.6
MD 45 7.7
ME 9 2
MI 25 6.3
MN 14 1.4
MO 4 6.4
MS 6 6.4
MT 2 2.9
NC 16 5.3
ND 4 1.5
NE 8 2.2
NH 6 0.8
NJ 72 3.7
NM 4 8.7
NV 5 5.9
NY 62 4
OH 9 4.5
OK 2 6.2
OR 17 2.2
PA 26 5.2
RI 44 2.9
SC 10 6.3
SD 4 2.6
TN 8 7.3
TX 6 5.4
UT 0 1.3
VA 16 4.4
VT 6 1.1
WA 17 2.7
WI 8 2.5
WV 4 4.6
WY 8 2.4
Yet we have other democrats just grating on the NICS check system being used for employment background checks are unreliable because of all the false positives that prevent employment. Then the false positives have to go thru a procedure to regain their good standing to get their employment, and guns.
ReplyDeleteWhich is it Mike? A good working system or a flawed one that presents false positives.
Same system used for both, but two different ways to look at it depending on your point of view.
It's a system that works extremely well. That's why Dom tried to steer the focus on the government's failure to prosecute people for lying on the form.
DeleteFor those politicians that are de-crying the false positives that the NICS check which employers use denying employment on the condition of passing the check will disagree with you.
Delete"The question of arresting and charging people for the paper crime of lying on a government form is a different issue and has little to do with the background check system's efficacy."
ReplyDeleteOk Mike,
So you believe that someone who lies on a government form is isn't a big deal if the government notices and doesn't allow the illegal activity, but then if the person who lies succeeds, you want to throw them in prison and take away all of their gun rights forever.
Presumably these are people who've already lost their gun rights.
DeleteThe point is Dom Raso tried to do a con job in that video. The 76,000 were stopped from buying guns, not just the 13 that were arrested for lying on the form.
Mike says,
Delete"Presumably these are people who've already lost their gun rights."
Your correct in that phrasing, presumably means not fact until proven. That's why there is a process to appeal a false positive. That's why prospective employment have to do the same appeals process to get the job they are applying for.
Texas, what percentage of the 76,000 were false positives, in your opinion?
DeleteI have no opinion on the percentages. Hard data is likely out there somewhere on the appeal process that is used to clear a wrongfully entered name. I just don't know how to find that but I am sure someone could.
DeleteThe reason I am using employment as an example is because of what I have seen on the news just a few days ago. The very same politicians (left handed do-gooders) that praised the NICS check and want to expand it for any gun purchase, turned right around and said how bad the system was for handling NICS checks for employment for rather critical jobs. In the SAME INTERVEIW!
Mike, it the same system. Exactly the same used for way more than just gun checks. If these people say the system is flawed for employment checks, then its also flawed for firearm checks.
Here is another thing, since check numbers are known and then adjusted for an ESTIMATE for firearm purchases and the rest left for OTHER purposes and none of the records of the check are kept and NICS does not know what the check was for, what percentage of the 76,000 were not for gun purchases but for OTHER reasons instead?
Why do you want Texas to give his Opinion on a percentage? So that you can say he's wrong, that your opinion is that the percentage is much lower, and that anyone who opines a larger percentage is lying?
DeleteMy point in asking for a percentage is that even if half of the 76,000 were for other than gun purchases or were false positives, the system is still working. Tens of thousands a year are denied. This is exactly the point that slick Dom wanted to move away from and focus on the government's failure to arrest those who lied on the forms.
DeleteAll you can say is that tens of thousands were denied the opportunity to purchase guns. Why they were denied, we don't know. And that's the kind of sloppy system that we're fighting. That much "descretion" shows that it's all about denying rights, rather than catching criminals.
DeleteGood stuff, Today, it's rare to see a runner at the courthouse. Increasingly, background checks are entirely computer-generated, sometimes matching records with the wrong person. These include inconsistencies in ages, addresses and even names.
ReplyDeleteemployment background checks