Analysis of Brady Campaign Statistics
- Thread starter denton
- Start date
- Status
oneounceload
member
What are you saying? All you did was link to something unknown
Please provide your view/opinion/analysis
Thanks!
Please provide your view/opinion/analysis
Thanks!
barnbwt
member
- Joined
- Aug 14, 2011
- Messages
- 7,340
"It uses county by county data, which is an excellent approach, but very labor intensive."
This was in reference to a study on suicides; we need something similar correlating gun laws (using some Brady or NRA metric) and various murders/rapes/robberies at the county level. Heck, even try to account for national income and racial trends, I bet you'd still find NYC, LA, and Chicago as horrible outliers inside their states, in both crime and gun control.
Gun control folks have always masked their failure to impact crime by laundering their statistics into larger data sets. It's the root of their argument that crime with guns in City X is due to guns coming in from outside the city where gun control laws are less, or from the next state, or the next country, etc.
County-by-county stats yields all sorts of interesting data that is obscured. My favorite was the comparison last election of the precinct totals by county across the US, followed by the same image distorted by population (denser = larger area). Looked like two completely different electoral outcomes, and got
me thinking about how urban areas have largely succeeded in overcoming the constitutional provisions set up to specifically limit their dominance.
TCB
This was in reference to a study on suicides; we need something similar correlating gun laws (using some Brady or NRA metric) and various murders/rapes/robberies at the county level. Heck, even try to account for national income and racial trends, I bet you'd still find NYC, LA, and Chicago as horrible outliers inside their states, in both crime and gun control.
Gun control folks have always masked their failure to impact crime by laundering their statistics into larger data sets. It's the root of their argument that crime with guns in City X is due to guns coming in from outside the city where gun control laws are less, or from the next state, or the next country, etc.
County-by-county stats yields all sorts of interesting data that is obscured. My favorite was the comparison last election of the precinct totals by county across the US, followed by the same image distorted by population (denser = larger area). Looked like two completely different electoral outcomes, and got
me thinking about how urban areas have largely succeeded in overcoming the constitutional provisions set up to specifically limit their dominance.
TCB
barnbwt
member
- Joined
- Aug 14, 2011
- Messages
- 7,340
"Since the data do not contain appropriate examples for a balanced two-factor design, which would separate density and Brady score"
Give you two guesses how "Brady score" and "population density" correlate, and whether it's stronger than suicide/pop density
. You may want to go back and 'reverse correct' (via anti-log? It's been a while since I did stats) the Brady/Suicide plot to show visually just how random the trendline becomes when controlled for population density.
Make a point of how population density has nothing to do with Brady's policy recommendations (to my knowledge) yet they promote the same curative to all regions of the nation. Make another point of how Brady recommendations largely* have nothing to do with suicide whatsoever in the first place, so the non-correlation makes logical as well as statistical sense.
"You'll find that in rows 63-85, they put "gun deaths" in as one of their inputs, along with "crime guns exported" in lines 87-109. These are not input variables. They are hoped-for outcomes."
Could you explain this further, please? Are they adding 'gun deaths' to raise, or lower, the Brady score of an area? I agree it has nothing to do with scoring the policy choices of an area** which is the whole point of the score, but perhaps explaining how and in what way it twists the data would be telling (I assume it slants it towards rural/low control areas, due to the disproportionate rate of unrelated suicide & the fact that 'gun deaths' are disproportionately suicides). I'd also put this tidbit ahead of the first chart; this adjustment will appear disingenuous if you reveal it after the eye-charts (heck, I'd put it up where you first refer to the Brady score). I'd also mention how it is a circular-reference, since "crime guns" are so-defined in part by virtue of the local area's gun control laws (assault weapons/NFA/concealed/etc.)
Apart from that, I'd say it's a pretty well thought-out report, devoid of spelling errors. To be honest, if you suitably mask your findings a little bit in the first half, this would be an interesting document to submit to media/scholarly areas, since it is quite damning & difficult to rebut (without getting into politics/demographics/etc.)
TCB
*In fact none do, but some of their schemes at least purport to reduce suicides by restricting access to firearms
**Assuming for the purposes of their argument that gun control reduces suicides, I suppose you could argue that actual reported gun deaths might be useful in indicating whether an area is actually implanting/enforcing their control scheme, vs. just giving it lip service. I think you'd want something much more nuanced than this adjustment to determine that, however (not to mention one shred of proof that said implementation was related to a demonstrable outcome)
Give you two guesses how "Brady score" and "population density" correlate, and whether it's stronger than suicide/pop density
. You may want to go back and 'reverse correct' (via anti-log? It's been a while since I did stats) the Brady/Suicide plot to show visually just how random the trendline becomes when controlled for population density.Make a point of how population density has nothing to do with Brady's policy recommendations (to my knowledge) yet they promote the same curative to all regions of the nation. Make another point of how Brady recommendations largely* have nothing to do with suicide whatsoever in the first place, so the non-correlation makes logical as well as statistical sense.
"You'll find that in rows 63-85, they put "gun deaths" in as one of their inputs, along with "crime guns exported" in lines 87-109. These are not input variables. They are hoped-for outcomes."
Could you explain this further, please? Are they adding 'gun deaths' to raise, or lower, the Brady score of an area? I agree it has nothing to do with scoring the policy choices of an area** which is the whole point of the score, but perhaps explaining how and in what way it twists the data would be telling (I assume it slants it towards rural/low control areas, due to the disproportionate rate of unrelated suicide & the fact that 'gun deaths' are disproportionately suicides). I'd also put this tidbit ahead of the first chart; this adjustment will appear disingenuous if you reveal it after the eye-charts (heck, I'd put it up where you first refer to the Brady score). I'd also mention how it is a circular-reference, since "crime guns" are so-defined in part by virtue of the local area's gun control laws (assault weapons/NFA/concealed/etc.)
Apart from that, I'd say it's a pretty well thought-out report, devoid of spelling errors
. To be honest, if you suitably mask your findings a little bit in the first half, this would be an interesting document to submit to media/scholarly areas, since it is quite damning & difficult to rebut (without getting into politics/demographics/etc.)TCB
*In fact none do, but some of their schemes at least purport to reduce suicides by restricting access to firearms
**Assuming for the purposes of their argument that gun control reduces suicides, I suppose you could argue that actual reported gun deaths might be useful in indicating whether an area is actually implanting/enforcing their control scheme, vs. just giving it lip service. I think you'd want something much more nuanced than this adjustment to determine that, however (not to mention one shred of proof that said implementation was related to a demonstrable outcome)
denton
Member
oneounceload....
The link is to a PDF that puts forth my analysis. It's really too long to condense into a post. In part, it's a response to this article, which is very persuasive if you don't have the background from my PDF. http://www.slate.com/blogs/the_slatest/2015/10/02/gun_control_by_state_tougher_laws_mean_fewer_deaths.html
barnbwt....
Yes, if you can go county by county it's a better analysis. But getting all the data is very difficult. So one question to answer is, can we use state by state numbers, without losing too much information? As it turns out, you can. And I'm just one guy working in my basement, on my own time. Still, I keep hoping for a big grant and a research staff....
Sure.
When you're trying to find a relationship between an output, such as suicide rate, and a set of input variables, such as purchase waiting times, state required background checks, and ease of getting a concealed carry permit, you can always be sure of finding a relationship if you put the output variable in as an input variable, too. That is what they did. The lower your suicide rate, the higher your Brady score. So when you check Brady scores against suicide rates, sure enough, you find a relationship. It's shown in the dictionary right next to "circular calculation".
Thanks, all, for the thoughtful comments.
The link is to a PDF that puts forth my analysis. It's really too long to condense into a post. In part, it's a response to this article, which is very persuasive if you don't have the background from my PDF. http://www.slate.com/blogs/the_slatest/2015/10/02/gun_control_by_state_tougher_laws_mean_fewer_deaths.html
barnbwt....
Yes, if you can go county by county it's a better analysis. But getting all the data is very difficult. So one question to answer is, can we use state by state numbers, without losing too much information? As it turns out, you can. And I'm just one guy working in my basement, on my own time. Still, I keep hoping for a big grant and a research staff....
Sure.
When you're trying to find a relationship between an output, such as suicide rate, and a set of input variables, such as purchase waiting times, state required background checks, and ease of getting a concealed carry permit, you can always be sure of finding a relationship if you put the output variable in as an input variable, too. That is what they did. The lower your suicide rate, the higher your Brady score. So when you check Brady scores against suicide rates, sure enough, you find a relationship. It's shown in the dictionary right next to "circular calculation".
Thanks, all, for the thoughtful comments.
denton
Member
That's discussed toward the end of the paper.
The Brady Campaign switched from letter grades to numerical grades. They also published the spreadsheet that they use to create the grades.
The Brady Campaign adds "gun deaths" in as part of each state's grade. But that's the output you're often looking for. If you put the output in as one of the inputs, you shouldn't be surprised to find a (spurious) correlation when you run the regression.
Since the Brady Campaign provided all the pieces that add up to the final score, I went back and subtracted out the outputs (there were two) that they had cooked into the inputs. That's the Deconflated Brady Score.
The Brady Campaign switched from letter grades to numerical grades. They also published the spreadsheet that they use to create the grades.
The Brady Campaign adds "gun deaths" in as part of each state's grade. But that's the output you're often looking for. If you put the output in as one of the inputs, you shouldn't be surprised to find a (spurious) correlation when you run the regression.
Since the Brady Campaign provided all the pieces that add up to the final score, I went back and subtracted out the outputs (there were two) that they had cooked into the inputs. That's the Deconflated Brady Score.
Sunray
Member
The Brady Bunch is well known for making stuff up. "Deconflated", for example, isn't even a word.
Thank you. I don't know if it makes any difference as people have a poor concept of statistics (I mean conceptually, I'm not even referring to the math!), but think we need to work harder on highlighting the facts over belief. Gun control relies strongly on belief or feelings and cold hard facts are a good solution to counter that, I hope.
barnbwt
member
- Joined
- Aug 14, 2011
- Messages
- 7,340
I'd lay hard on the circular reference aspect of their distortions early on in the report; that's about as big a story as anything else. I figured that was the type of game they were playing; why add the supposed outcomes to your policies as a means of describing them, other than to bake in a correlation off the bat? Didn't you also say they added 'found crime guns' or some nonsense as well? I would think that would further confuse the data, since gun laws are themselves responsible for making firearms into 'crime guns' in the first place (ie a 30rnd AR15 in CA is not illegal in Texas). If so, it seems more like the Brady folks simply don't know how to measure their policies' outcomes worth a darn.
"The Brady Campaign switched from letter grades to numerical grades"
Can't be that hard to defeat this whole premise from the get-go; assigning numbers to a diverse class of gun laws concocted across the country for a diverse set of problems is a fool's errand.
"Since the Brady Campaign provided all the pieces that add up to the final score, I went back and subtracted out the outputs (there were two) that they had cooked into the inputs. That's the Deconflated Brady Score."
I'd highly recommend adding this exact paragraph before your first infographic. That way, it is clear your adjusted numbers are both based on the Brady's own data, and adjusted by a logical and consistent measure that won't taint the outcome. Nothing more irritating than reading a paper on some sort of 'massaged' data that isn't clear on their reasons/methods for doing so
I'll bet you could even take it further, and turn the Brady correlation into a chicken/egg dilemma. Because gun laws have historically been passed for a number of simultaneous reasons (response to events, as well as demographic changes) which themselves are often the target of the policy, they cannot be untangled. Crime goes up so you pass a law, but unless the crime instantly responds to the stimulus (no evidence it ever does) you now have a law and a high crime rate in an area; correlation broken. Likewise, if crime goes down due to economic uplift in an area, laws may remain unchanged despite the shift; correlation broken.
Ultimately, these laws aren't resulting in lots of prosecutions, but rather about making it difficult to intentionally operate within the law. The result is that fewer choose to attempt to do so, but that really has nothing to do with criminal intent. Another interesting data experiment at the city or county level would be NICS checks (since I think county is reported) vs. Brady score. A stronger correlation between Brady score and gun ownership than with crime or suicide would be very telling about the real purpose of their policy recommendations.
TCB
"The Brady Campaign switched from letter grades to numerical grades"
Can't be that hard to defeat this whole premise from the get-go; assigning numbers to a diverse class of gun laws concocted across the country for a diverse set of problems is a fool's errand.
"Since the Brady Campaign provided all the pieces that add up to the final score, I went back and subtracted out the outputs (there were two) that they had cooked into the inputs. That's the Deconflated Brady Score."
I'd highly recommend adding this exact paragraph before your first infographic. That way, it is clear your adjusted numbers are both based on the Brady's own data, and adjusted by a logical and consistent measure that won't taint the outcome. Nothing more irritating than reading a paper on some sort of 'massaged' data that isn't clear on their reasons/methods for doing so
I'll bet you could even take it further, and turn the Brady correlation into a chicken/egg dilemma. Because gun laws have historically been passed for a number of simultaneous reasons (response to events, as well as demographic changes) which themselves are often the target of the policy, they cannot be untangled. Crime goes up so you pass a law, but unless the crime instantly responds to the stimulus (no evidence it ever does) you now have a law and a high crime rate in an area; correlation broken. Likewise, if crime goes down due to economic uplift in an area, laws may remain unchanged despite the shift; correlation broken.
Ultimately, these laws aren't resulting in lots of prosecutions, but rather about making it difficult to intentionally operate within the law. The result is that fewer choose to attempt to do so, but that really has nothing to do with criminal intent. Another interesting data experiment at the city or county level would be NICS checks (since I think county is reported) vs. Brady score. A stronger correlation between Brady score and gun ownership than with crime or suicide would be very telling about the real purpose of their policy recommendations.
TCB
denton
Member
OK, thinking about it for a minute, here's the paper in many fewer words.
1. The Brady Campaign, and many others, combine homicides/murders with suicides. These two things are not the same at all, and are products of different processes. It's not proper to add them. However, it might be the only way a writer can create a scary conclusion.
2. A state's Brady score does not predict its murder rate. If you do a scatterplot, with states Brady scores on the horizontal axis, and murder rates on the vertical axis, you find no relationship between the two. Running regression tells the same tale.
3. At first blush, it looks like Brady scores predict suicide rates. But this apparent association is probably not real. States with less dense populations have very much higher suicide rates. Population density is a much better predictor of suicide rates than Brady scores are. So the apparent association is probably due to the fact that less densely populated states also tend to have lower Brady scores.
4. The Brady Campaign uses "gun deaths" as one of the factors in creating its Brady scores. So if you're looking for a relationship between Brady scores and murders or suicides, you'll probably find one. If salt is an ingredient in your soup, you shouldn't be surprised of the finished soup tests positive for salt. Once you recognize this, and correct for it, and for the spurious association mentioned above, the Brady Campaign's statistics fall flat.
Since more strict gun laws do no detectable good, you have to ask, why would people want to tighten gun laws? As nearly as we can tell, that is an exercise in futility. On the other hand, about 1,000,000 times per year, people use guns to protect themselves and their loved ones. So having access to firearms does a lot of good in society. Clearly the good far outweighs the bad.
1. The Brady Campaign, and many others, combine homicides/murders with suicides. These two things are not the same at all, and are products of different processes. It's not proper to add them. However, it might be the only way a writer can create a scary conclusion.
2. A state's Brady score does not predict its murder rate. If you do a scatterplot, with states Brady scores on the horizontal axis, and murder rates on the vertical axis, you find no relationship between the two. Running regression tells the same tale.
3. At first blush, it looks like Brady scores predict suicide rates. But this apparent association is probably not real. States with less dense populations have very much higher suicide rates. Population density is a much better predictor of suicide rates than Brady scores are. So the apparent association is probably due to the fact that less densely populated states also tend to have lower Brady scores.
4. The Brady Campaign uses "gun deaths" as one of the factors in creating its Brady scores. So if you're looking for a relationship between Brady scores and murders or suicides, you'll probably find one. If salt is an ingredient in your soup, you shouldn't be surprised of the finished soup tests positive for salt. Once you recognize this, and correct for it, and for the spurious association mentioned above, the Brady Campaign's statistics fall flat.
Since more strict gun laws do no detectable good, you have to ask, why would people want to tighten gun laws? As nearly as we can tell, that is an exercise in futility. On the other hand, about 1,000,000 times per year, people use guns to protect themselves and their loved ones. So having access to firearms does a lot of good in society. Clearly the good far outweighs the bad.
"You can download the Brady Campaign's spreadsheet for calculating state gun law
strictness scores at http://crimadvisor.com/data/Brady-State-Scorecard-2014.pdf, on
page 6."
I'm still trying to follow the math. At least for my pdf viewer, pages 6 and following have verbal descriptions of the things the Brady folks consider, but there isn't a link to a spreadsheet or any information on exactly how those get turned into a score. There is a block that says 'Download full data set', but it's just a colored block, not a link.
I've googled for things like 'brady score spreadsheet' etc. w/o luck.
If anyone has the actual calculation details or spreadsheet location, could you post the URL?
thanks!
strictness scores at http://crimadvisor.com/data/Brady-State-Scorecard-2014.pdf, on
page 6."
I'm still trying to follow the math. At least for my pdf viewer, pages 6 and following have verbal descriptions of the things the Brady folks consider, but there isn't a link to a spreadsheet or any information on exactly how those get turned into a score. There is a block that says 'Download full data set', but it's just a colored block, not a link.
I've googled for things like 'brady score spreadsheet' etc. w/o luck.
If anyone has the actual calculation details or spreadsheet location, could you post the URL?
thanks!
denton
Member
Denton - got it, thanks!
Do you happen to remember the original url where you got this stuff directly off the Brady site?
From a 'chain of custody' POV, having it from, say, archive.org prevents anyone saying I'm making things up
(good work, BTW!)
Do you happen to remember the original url where you got this stuff directly off the Brady site?
From a 'chain of custody' POV, having it from, say, archive.org prevents anyone saying I'm making things up
(good work, BTW!)
vito
Member
The average person has no concept as to the validity of any conclusions based on any sets of data by any organization. All the typical citizen knows is when some media outlet treats the conclusions of a "study" as gospel. Even today we are still hearing the myth of "you are 43 times more likely to be killed in a house where guns are kept than in a house without guns" and similar distortions of the truth. Gangbangers are counted as "children". Drive by shootings a block from a school at 3 AM is counted as a "school shooting", etc, etc. When a valid analysis is done with data, such as John Lott has done on more than one occasion, the media and thus the public pretty much dismisses his conclusions by accusing him of being a shill for the gun industry and the NRA.
There's numerous problems with even their "crime guns exported". If they're pulling their "trafficked guns" data from the same place Bloomberg did then they're using ATF trace data: https://www.atf.gov/about/firearms-trace-data-2014. There's a couple issues with that though, and the ATF says it right across the top for every state. For one, they only trace to the state where the gun was originally sold at retail in most cases. So if someone moves or later legally sells the gun out of state, according to Bloomberg that's a trafficked gun. Another glaring issue is that the trace requests aren't even necessarily guns that were used in a crime [in fact, if you look at slide 6 for CA for example, you'll see that very few of the guns were actually used in a crime]. Then there's the problem that the ATF themselves state that the data is not representative at all, it's a rather random sample of trace requests by local LE.
denton
Member
happygeek...
Good info! Thanks.
We had a little breakthrough this morning. I figured out how to separate the Brady score influence on suicides from the population density influence. When you do that, the Brady score influence comes close to being statistically detectable. The population density influence is very strong, very significant, and 70X as influential as the Brady score.
Good info! Thanks.
We had a little breakthrough this morning. I figured out how to separate the Brady score influence on suicides from the population density influence. When you do that, the Brady score influence comes close to being statistically detectable. The population density influence is very strong, very significant, and 70X as influential as the Brady score.
Thanks for sharing. The junk science on the gun control side of this debate is astonishing. I would love to see some of these experts sit for a deposition with a well-prepared lawyer asking questions about their methodology. 'Twould be hilarious.
denton
Member
I've considerably re-done the paper, without nearly as much of the statistical details, and I've posted it as a blog.
Also, I've incorporated several suggestions made here on the board. Thank you all for your comments.
You can review the result at:
https://notsofast2015.wordpress.com/
Also, I've incorporated several suggestions made here on the board. Thank you all for your comments.
You can review the result at:
https://notsofast2015.wordpress.com/
barnbwt
member
- Joined
- Aug 14, 2011
- Messages
- 7,340
"For one, they only trace to the state where the gun was originally sold at retail in most cases."
Very important point that they are only talking about federal 'crime guns' and not state-level offenses; statistically, the states with most gun sales will rank highly here, since so many 'crime guns' are in fact stolen. Thus this is yet another self-referential/self-serving data distortion which solely serves to inflate the 'Brady score' in states with lower per capita firearms sales (a fact that I'm sure does cause them to look more kindly on a state, but that they would never admit to publicly). Otherwise, a state like California and all its assault weapon laws would rank higher per capita due to the labyrinth of laws and rampant black market they brought to life.
TCB
Very important point that they are only talking about federal 'crime guns' and not state-level offenses; statistically, the states with most gun sales will rank highly here, since so many 'crime guns' are in fact stolen. Thus this is yet another self-referential/self-serving data distortion which solely serves to inflate the 'Brady score' in states with lower per capita firearms sales (a fact that I'm sure does cause them to look more kindly on a state, but that they would never admit to publicly). Otherwise, a state like California and all its assault weapon laws would rank higher per capita due to the labyrinth of laws and rampant black market they brought to life.
TCB
- Status
