This is a false statement, arising from ignorance, I’m sure. From the Kellermann et al paper:
You claim to have read the paper, so you must have read this. Yet, you then claim that those variables were not controlled for.
Which part of the above quote do you not understand? Be specific, and I’ll help you. Otherwise, acknowledge that you made a false statement.
This is another false statement - see Item 1 of Post 623. You would discount the results if the covariates in the model led to instability due to low cell counts (or other model diagnostic concerns). Please stop making so many false claims.
You seem very confused about case-control designs. Perhaps you’re thinking of random assignment designs?
Case and control groups will always differ, unless you’re looking at a condition in which there is only one mechanism that causes that condition. If you study cancer, for example, and match cases with cancer against controls without cancer, you’ll get differences on diet, obesity, alcohol consumption, tobacco use and other factors.
It isn’t sorcery that allows us to include a measurement of asbestos exposure to the data set to identify the relative contribution (even in the face of differences on other factors) of exposure to asbestos in the risk for cancer! Imagine how hampered science would be if we could not control for differences on more than one factor at a time.
You must know this, so this is a total failure of logic.
Look at this guy! He really doesn’t get that he’s saying the same thing. ‘I don’t dispute that that the study shows that people who had a gun in the house were twice as likely to be killed, but you can’t say that people with a gun in the house were twice as likely to be killed.’
Yes, I read the paper. Let’s parse the 2 step process:
[ol]
[li]models containing closely related variables (such as those describing the use of alcohol in the home) were constructed to identify the variable or variables in each set that were most predictive of whether the household in question was a case or a control household.[/li][li]Next, a model that incorporated the variables selected in this initial step was constructed to select those that remained significant after we controlled for the effects of the remaining variables in the model.[/li][/ol]
It should be clear that my reference was to the multivariate analyses, since I quoted from that section directly - is this an area where you were struggling with reading again? How do you reconcile that with the statement from the study:
This is in respect to the multivariate analysis that is identified in Table 4, the source of the 2.7 figure - you know, the one being discussed. Clearly you’re not saying there was multivariate analysis where the other four factors above (drinking and medical attention due to fighting) were controlled for, right? Talk about baffling with bullshit. From that same paragraph about multivariate analysis:
Of course, in the section you quoted, it states:
So all of those other criteria did not significantly alter the adjusted odds ratios of the items in Table 4. Essentially that those other items that I identify that were large disparities between case and control were just non factors. If you want to call that controlled for when the final model only includes the 6 variables, that would at least be consistent with your penchant for misleading summarization. I won’t belabor this because it doesn’t really make a difference to my point that the nature of the case and control groups were wildly different, and that the nature of those characteristics that place people at greater risk could also lead them to acquire firearms which would lead to a positive association between homicide and firearm ownership even if that ownership had no impact on the later homicide.
Let’s see, here’s what you said in response to #1 in post 623:
Compare that to the above. In post #623 you stated that problems would not arise unless there were cells with no observations. But just above, you state that there would be problems (discounting the results) if the covariates led to instability due to low cell counts. Those are different statements - is there one you’d wish to go with? Do you consider the question answered if you answer each time differently? Is that like calling 71% “all” like I identified in post #638? In my example from post #638, if the 3 time felon case group was compared to the no arrest suburban housewife, you’d find no problem with that? They could match on the 4 criteria used to find control samples.
And here is where your rush to quip seems foolish since I acknowledge in post #625 that multivariate regression can be done even with great differences between two levels of other variables. I question the value when the analyses is done so poorly.
See how this works, you accuse me of making two false claims which are easily rebutted. I have accused you of misleading quoting (post #589, 591, and 622) as well as misleading summarization (post #625 and 638) to make your position look better. Personally, I would find that accusation particularly objectionable and either hastily rebut the notion, or if in fact I did engage in that sort of shoddy debate tactic I would retract. You seem fine with it, which informs how to interpret the rest of your conclusions.
In any event, there is even further analysis why the 2.7 figure is crap which will take longer to parse so I’ll leave that for another time.
If you did that, there would be big cultural changes, perhaps including civil war. I agree that the studies don’t say anything about that.
I’m on the side of discouraging gun ownership. But I admit that gun owners do differ from others in subtle and relevant ways that were not addressed in the case control study. Here’s the science:
And here’s the red-meat take:
http://csgv.org/blog/2015/psychology-gun-ownership/
The question I would ask and answer is: Does the preponderance of the evidence point to, for people on the fence about being armed, more or less safety when bringing a gun into the household. According to the friends of the guns, or at least the ones I know, an armed society is safer. But the state by state evidence argues against it:
Those who don’t care about suicide – most US gun deaths, outside some large cities, and Louisiana, are suicides – may dismiss this. So be it. But I don’t see how the dangerous parts of the US, like Alaska, Wyoming, and Montana are so much more filled with anger and depression than the safer places like New York and Massachusetts. It seems to me more plausible that there also are impulsive angry people in New York who would have killed someone (probably themselves), but didn’t because of the delay and difficulty involved in getting a gun.
Bone is just desperately arguing from a position of sheer ignorance. His sweaty desperation in trying to suggest I’m in error when I talk about zero cells versus low cell counts should be an embarrassment. It’s a bit like the stuttering lawyer in My Cousin Vinny, taking desperate stabs during cross-examination.
Remember that his obsession has been that homicide cases differ(!) from non-homicide controls on factors. Upon learning that there isn’t a “sense” as to when there are problems, but actual specifics, he first tells a falsehood about the issue, and then desperately grasps at something he doesn’t understand. Zero cells are a thing - google it. There would also be POTENTIAL concerns if the EXPECTED cell count is less than five.
How does one explain a Coke bottle that falls from the sky to a man with no contact with the outside world? Pure sorcery that must be destroyed!
Is there anyone else who understands stats and can take this on? Teaching these people here takes up a lot of time. Now that I say that, it dawns on me that everyone else is probably too smart to bother even trying in the first place.
It’s also darkly comical to note that all this desperate straw grabbing is an effort to tear down one study that has been repeatedly replicated since then.
So how do you explain your previous attempts to portray the studies as having some sort of causal link?
If you are now willing to concede that the studies do not imply a causal link between the presence of guns and murder (at least not at the substantially increased levels that the studies identify) then I am happy to accept your concession but please don’t act like you weren’t presenting this study as establishing some sort of causal link between the presence of a gun and murder.
How many other gun control folks have latched on to what you said and made statements to the effect that you are actually putting your family at over twice the risk of getting murdered when you bring a gun into the home? All because you keep implying that the gun is the source of the increased risk.
Credibility is earned. I suspect you once had some, coming clean and admitting that you were portraying these studies as having some significant probative value when they have little to none would be a good first step towards getting some of it back.
No studies prove cause. The hypothesis that guns play a casual role to increase risk has not been contradicted in any of the studies yet. The fact that the risk remains after controlling for multiple factors in multiple studies is totally fucking “probative.”
ETA: Credibility is earned. Credibility is also lost when you engage in desperate antiscience handwaving AND conspiracy theorizing AND arguing from rank ignorance.
I’m not sure whether my number there is the same as what you are thinking, or otherwise.
How many any kind of folks have a strong preference for the death of their family to be by murder rather than suicide? Hopefully zero as well. My preference is that they all live to old age, so I would focus on rates of homicide (less common) plus suicide (more common).
I’ve never brought a gun into my home, but I have brought in fairly serious-looking kitchen knives. Here are the risks of stabbing vs. shooting where I live:
The same is true if you try to force a million people to smoke.
My point was that these studies have a predictive value with smoking that they do not have with guns. If I force you to smoke then YOUR chances of getting cancer have gone up. If I force you to keep a gun in your home (perhaps because you live in an area permeated with gun culture so you feel pressure to own a gun), you have not similarly increased your chances of someone in your home getting murdered.
The things i think they are not controlling for are pretty fucking obvious and perhaps they are not easy to measure but until they are measured you can’t say anything like what Hentor has been saying with any degree of confidence. You are just guessing.
I’m not sure how those cites are useful here.
I don’t know that an armed society is a significantly safer society but judging from the fact that our suicide rate is pretty fucking average for industrialized nations (where they seem to manage to kill themselves pretty easily without the use of guns) its hard for me to attribute our suicide rate to guns. So no, I don’t really consider suicide very much in the gun debate until and unless someone can show me some evidence that our long term suicide rate would be significantly lower in the absence of guns.
And I might agree if the states with the highest gun murder rates were also the states with the highest overall murder rates but it turns out that overall murder rates don’t really correlate that well with gun ownership rates.
I take it as a given that suicides will be more commonly committed with a gun in a place without a lot of tall buildings but lots of guns and without suicides your correlation kind of evaporates.
Dude, you just learned about zero cell counts and their meaning in the context of logistic regression. You were introduced to the issue of expected cell counts as well, and although I did not explain much about what they are, I did teach you that low expected cell counts (conventionally below 5) are a concern. I’ve helped you improve your grasp of covariates, variable selection techniques, case-control analysis, odds ratios as an index of association - just to name a few new pieces of knowledge you now have that you previously did not. It takes some fucking gall to accuse me of offering content free posts.
It comes back to the sense that this is sorcery to you - if the content is so far out of your reach, it must seem pretty ethereal.
It doesn’t matter how many times you replicate the results when you use the same methods and criteria. It doesn’t matter if you control for “multiple” factors when you fail to control for some obvious factors.
Using your approach, I could say that living in a trailer park triples your chances of getting lung cancer or emphysema because after correcting for things like race, age, gender, diet, and pretty much everything except cigarette smoking; and it turns out that living in a trailer park really increases the chances that you end get lung cancer and emphysema. In fact other folks have done dozens of similar studies showing this exact same relationship between trailer parks and lung cancer/emphysema.
The studies all say the same fucking things and none of them undermine a potential causal relationship between trailer parks and lung cancer/emphysema. NONE OF THEM contradicts the hypothesis that trailer parks play a causal role in lung cancer and emphysema. The studies all control for a shitload of variables so I can now confidently declare "living in trailer parks triples your chances of getting in lung cancer or emphysema.
My point is that these studies would be a lot more probative value if they accounted for some pretty obvious things. But they don’t. We keep asking you how this gun presence translates into more murders and it might as well be magic as far as you are concerned.
Gang members are ten times more likely to own a gun than their peers. Gang members are 20 times more likely to get murdered than their peers. This seems like a dramatic enough confounding factor that it really ought to be accounted for. The Kellerman study controls for dozens of variables and yet leaves out things like gang membership. And that is just ONE reason why the Kellerman study and all the tohers that fail to take gang activity into account do not have very much probative value.
Do you get it now?
Of course I’m doing none of those things but true to form your arguments are short on substance and long on insults.
And yet you offer content free posts. I’m not sure what level of arrogance it takes to assume that it was you who introduced or taught anything. Why would you make that assumption? It would be like me claiming to have introduced you to counting - except that you clearly made an error there and you’ve demonstrated no such thing on my part.
You realize that this is not the only thread where this study comes up, right?
Guns are not a necessary element of suicide. The folks in Korea and Japan seem to be quite proficient at it despite their utter lack of access to guns.
If you want to make an argument that guns cause suicides then go ahead but you have to explain away the fact that our suicide rate is absolutely average despite being up to our eyeballs in guns. Sure there will be differences at the margins, but I think the substitution will take care of most of the differences over the long run.
Are you under the impression that the study is saying that having a gun in the home means you get murdered by THAT gun? The study specifically says that this is NOT what they are saying. In fact most of the murders are being committed by something OTHER than a gun that was kept in the home.
Only 14% of cases where we know the origin of the gun used in a homicide in the Kellerman study were identified as the gun kept in the home. And guns were the the only method of homicide in the study (we might also infer that the guns with unknown origin are more likely not from the home).
Your comparison with knives seems irrelevant in this context. If I had said that “knives kill too” then your point might be relevant.
The studies DON’T use the same methods and criteria. We’ve already covered that.
Here’s what the literature says: this is an effect particularly associated with family or intimate homicide. It is independent of prior arrest, criminal behavior, prior violence in the household, illicit drug use, alcohol use, and SES among other factors. It is found when restricted to legal purchasers of firearms.
Wha aspect of gang membership is independent of all of these factors, is found among legal firearms purchasers, and would be particularly related to family homicides?
I know the shit you do must seem pretty important to you but its not rocket science, anyone can get a general gist for what these things mean. Its not like you’re trying to explain why the fact that the speed of light is constant regardless of the observer’s speed leads us to the special theory of relativity. There might be important distinctions between things like odds ratios and rates and correlation when you are talking to your colleagues but for purposes of debate on this board, the differences really aren’t important enough to make or break an argument and your pretending otherwise just makes it look like you are appealing to authority. It looks like you are trying to baffle us with bullshit because you can’t dazzle us with dexterity. IOW, bring an argument when you come back.
You are at once saying that this stuff is easy to understand and also that understanding it doesn’t matter.
I totally agree with the first part. We’re talking about fucking correlations, odds ratios, and regressions. This is introductory level shit. Yet, you don’t understand, and you keep saying the same stuff. So it is necessary to this discussion.
Again, if you were to say that the correlation was .6, and I replied “you’re confusing correlation and causation,” would it not seem like I was fundamentally misunderstanding something?
And Bone, I am assuming that these concepts are new to you because someone who understood them wouldn’t say the things that you are saying, unless they were being deceitful.
If I said “the correlation between A and B was .6” AND I said “therefore we should stop doing A because B is bad” and you replied by saying “you’re confusing correlation and causation” then would it STILL seem like you are misunderstanding something? Because that is what you have been doing with this study. You have been presenting it as evidence that we should take at face value that if you bring a gun into your home you are making your family significantly less safe. You are doing more than just stating an observed association, you are imputing causality.
And there also are tremendous state by state differences in gun ownership – enough to explain a large portion, perhaps most, of the Japan-style suicide rates in the strongest gun culture states.
I’m quite open to the idea that Japan’s historical tradition of honorable suicide is a big factor. But why do Montana, Alaska, Utah, New Mexico, and Idaho all have higher-than-Japan suicide rates despite NOT having an honorable suicide tradition? Instead, what those states have is a tradition (not sure how old this tradition is, but it’s a tradition by now) of keeping guns around the home and person, so that when impulse to self-destruction arises, the most effective tool for said destruction is readily at hand.
Culture matters a lot. It explains why some states have a high suicide to murder ratio (Wyoming), and others have a low suicide to murder ratio (Louisiana). Guns explain why both states are, by first world standards, unusually dangerous.