And the Junk Science Begins

I am by no means an expert on statistical analysis, but I’ve probably had more training in it than most people, and this looks like a crock of horse shit to me so far. I’m not sure what the graphs they are publishing here even mean, or even how they were arrived at. At best I can tell it’s boiling down self-defense situations to something extremely simplistic by people who are not subject matter experts in that field. Self-defense can’t be boiled down to a probability calculation. Because it involves people, it’s not something you can model mathematically.

I also don’t understand how you can treat gun violence as an epidemiological issue when it’s not a disease. I think you are already making an awful lot of assumptions if you’re starting out with that approach. I see no good coming from this, which is why we need to continue to fight funding to this garbage.

UPDATE: The paper is here. This is the sentence which can clue you off that it’s junk science:

This debate cannot be settled satisfactorily by verbal arguments alone, since these are often driven by opinion, and lack a solid scientific backing. What is under debate is essentially an epidemiological problem: how do different gun control strategies affect the rate at which people become killed by attackers, and how can this rate be minimized? This question can be addressed with mathematical models that describe the interaction between a criminal shooter and one or more people that are the target of the shooter.

Emphasis mine. That sentence right there ought to raise alarm bells. If complex human behavior and interactions could be mathematically modeled, we’d be quite comfortable with the idea of armed robot police. We’d have pretty good artificial intelligence. Even for relatively predictive human activities that can be done relatively well by computers, like driving and flying, we’re still not at the point where we’re comfortable turning everything over to machines.

The results then clearly depend on the assumptions underlying the model, and this is very important to keep in mind when reading this paper, or any paper that deals with mathematical models in biological and behavioral sciences.

Yes, certainly, but you can bet that the media and policymakers will completely ignore these caveats and dive straight into the conclusion, in order to give their own preconceived notions the air of science, and only some stupid gun owning neanderthal with barely any education would argue with science.

UPDATE: One other severe criticism I have with their model is that it only considers homicide. Do we not care if someone, I don’t know, gets their head beaten against a sidewalk and suffers permanent brain damage? Do we not care about whether someone can avoid a lengthy hospitalization and recovery from, say, multiple stab wounds? Do we not care about rape? (Why do they so hate women?) Do we not concern ourselves with how often these kinds of horrors can be avoided by merely drawing the weapon on the attacker? You can’t just boil down human violence to homicides. That’s highly simplistic and completely invalid.

UPDATE: Joe Huffman has more to say in regards to this. I’m still getting caught up with everyone, having effectively taken Wednesday through Friday off blogwise, so I’m running a bit behind on what folks are talking about out there in the gun blogosphere.

27 thoughts on “And the Junk Science Begins”

  1. These people are shills, and their latest “research,” is nothing more than shilling for Obama. The fact that constitutional rights, are an obvious societal benefit, is not lost on these them. Rather, they have to come up with all sorts of fraudulent devices, with which to try and prove, the illogical. Bombard the public with propaganda in disguise as research. That they fail to mention the why, behind a lack of funding for such “research,” is practically a confession. Despite the fact that there is actually peer reviewed research. They don’t like it when the truth smacks them in the anti-gun face

    1. A big problem with research like this is that you are still looking at statistical aggregates. It doesn’t tell you much about any individual situation. In other words, you can’t suggest that because police only hit their target in self-defense situations 1/4 of the time, and criminals hit say 1/2 of the time, that it therefore makes sense to disarm cops (that’s not the actual stat, but just to illustrate). Nor could you use that fact to make predictions about the outcome in individual gunfights. You can put together models like these folks have, and I can’t, just reading through it, find any serious flaw with their math. It’s just that I don’t think their equations will tell you anything useful.

  2. Yes, because a mathematical model equals science.

    What the media and policymakers ignore is that said models are only scientifically useful if they are *predictive*.

    That is if they can acutally match and predict real world phenomena.

    And that doesn’t even get into that basic human rights trump science. Like say if you could “prove” certain neighborhoods produced a higher proportion of violent criminals.

    What then? Would denying people rights based on where they live, based on what an equation said, be okay? Just because it’s *science*?

    That they see violence as the same thing as gun violence as the same thing as guns as the same thing as a disease shows they’re operating from a deodand frame of reference.

    And really that tells you all you need to know about how “scientific” they are.

    1. I haven’t looked into their study yet, but common practice in predictive analytics is to train the model on 2/3rds (or some other portion) of the data and save the other third to test how well the model predicts. I have a feeling they didn’t do that.

  3. In response, University of California, Irvine professors Dominik Wodarz and Natalia Komarova created the first mathematical model to measure how legal gun availability impacts firearm-related homicide rates, published today in PLOS ONE. [my emphasis]

    Well, there is their problem. You know it’s junk science when their model says if gun control results in someone being stabbed to death instead of shot to death, it’s all good.

    Regarding your update, there’s that old saying of “you can’t hide a body” making homicide statistics more accurate than other violent crimes. That makes it a good reason to separate it, but maybe not to limit the study to homicides (though funds might be a reason). Speaking of which, the article had more whining about blocking federal funding for this type of junk, as if things can’t happen unless big government makes it happen. Irritating on so many levels.

  4. Why the the new national propaganda campaign against lawful self-defense? I see two objectives the enemy is pursuing.

    The first objective is a fallback position against the 2nd Amendment. The anti-gunners lost the debate about the legal and historical true meaning of the 2nd Amendment, so now they are hoping to undermine the basis of the amendment by trying to convince the public to give up self-defense. This is a battle even more lopsided than the legal/historical fight and one the enemy is doomed to lose.

    The second objective is another form of playing the race card. They are trying to boost support among the hard base of the Democratic party. I am seeing a truly ugly synergy that I believe is deliberate strategy. They are trying to combine gun-control politics with racial grievance politics. Each leaning on the other in a bastard synergy, but gun-control is most definitely the junior partner here.

  5. Models NEVER equate to reality… And they have no real data to compare with at this high a level… You are correct, you CANNOT model human interactions correctly, simply because we ARE humans, not machines!

  6. Statistics mean nothing to me. In fact, the social sciences in general are mostly all b.s. You can’t predict or control organic human behavior with abstract models. But over-educated liberals have been trying to do this ever since the Vietnam War.

    I’ll take logic and my gut instinct over stats & social science every time.

    1. And you say that as a trained social scientist or based on something you read on the internet somewhere? You assume that social scientists aim to predict when, in fact, it is generally our/their aim to describe. Correlation is not causation and any social scientist that tries to tell you that they can predict future outcomes based on a social model (no matter how large the n) is a fool or a liar. You also assume that social scientists are all over-educated liberals. My classmates were all active duty field grade Army officers or retired officers… not a liberal amongst us.

      I won’t argue with your desire to follow your gut. If that works for you, great.

      As to the paper, there are numerous methodological flaws (n is too small, model is derived from a model developed to describe combat, ignores non-fatal engagements, cites both Kellerman and Kleck…) but I don’t think it’s as bad as some of the comments here indicate. It’s not great but it’s a lot better than most of the surveys/studies I’ve read. The money shot for me was, “Of course in the real world a reduction in gun-related (and other) homicides would require improvement of the living and work conditions and education of underprivileged populations.” If you read past the underprivileged code there you’ll see that the authors understand that this is far more than an issue of the number of guns somehow correlating with homicides.

      I think that the authors of the study set out to do some real modeling but were hamstrung by their own assumptions (note the number of citations concerning crime with relatively few related to self-defense) and the paucity of research and data concerning what the readers of this blog undersand to be the legitimate use of firearms for self-defense. In short, they likely learned from liberal professors that guns cause crime so that’s where they started their research. Along the way they recognized that there is a lot more to this than mere gun control legislation.

      If you look at this in comparison to most of the other craptastic “research” out there (Kellerman!) it’s not that bad. Consider the conversation many of us have had with a friend, coworker or loved one where their starting assumption is that guns are icky and spontaneously murder whole pediatric oncology wards. With a few well-constructed arguments and citation of relevant facts many of us are able to alter that perception. The same thing can, and does, happen in academia. I wish I could see the first draft of this paper sent to the peers.

      This particular subject is my passion. The topic of my dissertation, though not yet accepted by my committee (a little trip to Afghanistan has thrown a wrench in my plans), is an examination of the junk science related to gun control. It will be years before it gets published now but there are definitely social scientists interested in exploring all of the data and presenting accurate statistical descriptions of this complex problem.

      1. I say this as a Humanities Ph.D. No offense, our fields are rivals, and yours is clearly winning, so I had to get a dig in somewhere.

        Good luck finishing your dissertation. That is a study that definitely needs to be done.

      2. J,

        If you look at this in comparison to most of the other craptastic “research” out there (Kellerman!) it’s not that bad.

        The above sentence is from the online dictionary entry, used to illustrate “damning with faint praise”. :-) Right up there with “less toxic than Sarin!!!”

        Seriously, though: I get your point, and it’s not a bad one. But (and you had to know ‘but’ was coming…) if it’s not useful for prediction, it’s really not useful for helping formulate good policy, is it?

        1. It is, indeed, faint praise. My larger point is that it is, at least, not virulently anti-gun nor is it using plainly unscientific methodology as just about everything spewed by the VPC and its ilk. I could be wrong but I think the authors could be convinced to open their aperture a bit.

          Ideally, there would be some scientific method of predicting human behavior. Failing that (which we are obviously doing) description can work pretty well. For instance, there is no way to tell what rioters will do, yet every major police department in the country has policies regarding riots as they have seen enough of them to describe them generally. As a researcher, if I can accurately describe the instances where illegal gun violence takes place I have some hope of providing policymakers with enough information to write good policy.

          I’ll illustrate one more way. There is little doubt that drunk drivers are dangerous. Scientists have described the circumstances where drinking and driving is likely to cause accidents, injuries and fatalities. For the most part, a BAC of >.08 is considered unacceptably impaired for motor vehicle operation. Though there are “functional” drunks that routinely drive with a BAC well in excess of .08, scientists have described human behavior accurately enough to provide policymakers with a pretty good idea of what should be legal and what shouldn’t.

          All that said, I fear that the description of those most likely to commit crimes with a firearm is such that policymakers will be unwilling to write good policy in response. That’s why they focus on the scary gun rather than the scary person.

          1. The problem is that this kind of thing can be too easily manipulated by the unscrupulous. For instance, in this paper, they are honest about the limitations. There have been plenty of cases where the media and policymakers ignore that bit because it doesn’t fit with their political agenda.

            When our side has the advantage in energy, and in numbers, I’m actually more comfortable waging the fight on those terms. Ultimately I’m not really all that concern with what the science says, because that’s going to deal with aggregates. It’s an individual right, and that will trump whatever the societal cost is of recognizing that right.

    2. “But over-educated liberals have been trying to do this ever since the Vietnam War.”

      And you apparently believe that over-educated conservatives, being aware of a successful winning tactic, haven’t been doing the same thing — to us?

      1. There haven’t been real conservatives in the U.S. in a long time. All those in power are big-spending imperialists.

        1. Too true, but don’t go telling that to all their puppets, who have been told THEY are the true conservatives, and believe it to the level of being a religion.

  7. By any chance do you suppose these would be the same math geniuses who did all the math models for the global-cooling/global-warming/climate-cooling/climate-warming/we’re all going to die if we don’t go back to the stone age modeling?

  8. I am probably educated to a similar level as Sebastian in terms of the mathematics of statistics, but the more important issue than the math, strictly speaking, is the entirely human, psychological tendency to see mere “correlation” as scientific “proof” of causation.

    I want to keep this issue-neutral, so I won’t cite the issue, but the other day someone sent me an article from a generally credible source, in turn citing a paper by a professional organization, that in turn cited almost a couple dozen independent studies, that had found that people who did “X” were more likely to later suffer mental illnesses. The language was such that it implied it was “proven” that doing “X” could CAUSE a person to subsequently be diagnosed with a mental illness. It did not address at all, whether people who already suffered from incipient mental problems might be more likely to do “X.” But everyone who hoped to prove that doing “X” had downsides — presumably including the editor of that article, and the authors and sponsors of some of those studies, who should have known better — went forth claiming that doing “X” might lead to mental illness.

    It is another form of “confirmation bias,” something we all are guilty of.

  9. Mathematical modeling is what insurance companies use to determine rates. It does have validity for aggregates but not for individuals.

    1. Exactly. But, none of us as individuals ever wants to accept that we may be behaving as predictable herd animals, even when our presence in an aggregate population suggests that is PROBABLY true.

Comments are closed.