Bone, buddy… I mean, apparently the technology to assess the degree of association among more than 2 variables at one time seems like voodoo magic to you, but it’s prettty old school stuff. I can see you’re going to fight this one to the death, but I have to say that science would not have progressed very far if we weren’t able to use this sorcery.
I just wonder what you must think of other research that uses even more wonderous forces than this thing you call “regression analysis”. Like behavior genetics! Woo0ooo0ooooh! Latent variable analysis! WooWoooo0oooo0!
Perhaps if you engaged in less psychotic laughter you wouldn’t ignore so many of your argumentative deficiencies. Let’s highlight the things you’ve ignored in this thread thus far (this leaves out your misleading quotation of source material, related to your error I identify in #3 below):
[ol]
[li]At what point would such a disparate group [in a case control study] render the results meaningless in your mind? [/li][li]If an unarmed person starts trying to kick in my door, and through the window they see me display a firearm so they run off, no shots fired, no police report - do you consider that a DGU? If yes, how would that be captured by any study you’d accept? If no, why not?[/li](for this one, you somewhat answered, though you weaseled out on the criteria and implied it would only count as a DGU if there was a police report. You completely ignored the second part.)
[li]Research on the other hand [at the CDC], has always been permitted, contrary to what you previously asserted in error. Care to walk that back or are you going to omit discussion of that one as well?[/li][li]And while I did not directly ask, I referenced the disparity in your acceptance of surveys/interviews based on the results of those inquiries, why is the methodology acceptable in one scenario but not the other?[/li][/ol]
Look, your strategy is clearly to try to baffle with bullshit. That’s why your posts get longer and longer as you start failing in a debate - you start introducing all kinds of extraneous bullshit so that later you can claim people are avoiding your shit.
[quote]
[ol]
[li]At what point would such a disparate group [in a case control study] render the results meaningless in your mind? [/li][/quote]
It is not in my mind. It is statistics. Please acknowledge that you know this as an entire field of study, with well established application.
Problems would not arise among dichotomous covariates unless in crosstabs (tabulations of the values of two variables against one another) there were “zero cells” - cells with no observations. For example, if there was nobody in the Kellermann study who had both no “history of illicit drug use” and also “did not keep a gun in the house”. Stop and think for a minute about what you’re claiming. Does it really make sense to you that mulitvariate regression analyses could not be done if there was a 30% difference between two levels of another variable in the model? Really? That is completely laughable and betrays not only ignorance of statistics but of logic as well.
See, here you recognize that you’re embarrassing yourself about simple statistical techniques, so you bring back something you admit I’ve already addressed in order to try to distract. Do you admit to being full-on ignorant about the statistical complaints you’re tryng to make? If not, don’t try such pathetic distractions.
[Quote]
[li]Research on the other hand [at the CDC], has always been permitted, contrary to what you previously asserted in error. Care to walk that back or are you going to omit discussion of that one as well?[/li][/quote]
Like most Conspiracy Theorists, you believe you have special knowledge that most people don’t have. Here’s what normal people know, from a google search of “gun research funding.”
Why do so many news outlets talk about the ban on funding for gun violence research? IT’S A VAST CONSPIRACY! They are all in on it too!!!
I have answered this. More bullshit distraction techniques, and a complete falsehood as well. I told you that the lack of replicability along with demonstrated problems of telescoping, magnification, implausible reports and evidence that more than half of reported DGUs are illegal. You responded by claiming that the people involved in demonstrating this were in on the conspiracy.
So, you baffle with bullshit, make false claims and endeavor to distract, and then turn around and suggest I am avoiding the issues. Poor form, little guy.
Unfortunately, it does take more effort to call out your bullshit than it does for you to vomit it out. If reading is so difficult, perhaps you can work on your counting? Do you have trouble with both of these skills? I had thought it was just the counting.
What do you think I’m claiming? Yes, multivariate regression can be done if there is a 30% difference between two levels of another variable. If there were control groups with 30% variance between them and the case, and also control groups that more closely aligned with the case group, it’s obvious to me that would be superior for the purposes of analysis. Yes, multiple variables can be controlled for, but the more that is required to be controlled for, and with greater magnitude, it makes the results less meaningful.
Look at Table 3. In addition to the factors I already mentioned in post #620 comparing the case subjects to the control subjects, here are some others: (all figures below are case vs. control)
Drinking caused problems in the household: 24.8% vs. 5.7%
Any household member had trouble at work because of drinking: 9% vs. 0.8%
Any household member hospitalized because of drinking: 11.4% vs. 2.3%
Any family member required medical attention because of a fight in the home: 17.3% vs. 2.1%
And your predictable response, that was all controlled for! If Kellermann simply used a control group that was more similar to the case group this criticism would be mooted. Given Kellermann’s other poorly conceived studies (relying on police reports to indicate victim weapon use when the reports were not designed to capture that data for example), I’m skeptical if he isn’t primarily agenda driven. He’s tainted.
Again with the misleading summaries - I guess it’s like your thing. I admitted you addressed a portion of the question - but ignored the largest part. I see you failed to address it. Are you ignoring it on purpose? If there is no police report, how would that be captured in any study you’d accept?
Maybe because it makes better headlines? Your mistake would be to accept these as true without going to the source material. It was referenced in a story you quoted - oh wait, you edited out the part that indicated what was actually restricted was advocacy. You asked previously if research wasn’t banned why there was an executive order necessary to initiate research? But the thing is, an executive order cannot overrule the will of Congress in this arena. The prohibition they passed [not the NRA, maybe watch some Grade School Rock to help you out in this area] was unchanged, yet the research was conducted! Go figure - it’s like research was never banned. And wait, here’s the research that was directed to be conducted. Do you think that the president can overrule the will of congress in this arena?
I guess misleading summaries are really your thing. I think you’re the only one going on about some conspiracy. Your “evidence” consists of anonymous judges funded by Joyce. Let me know if you’d accept research conducted by the NRA - that’s rhetorical.
But good on you for trying you little moppet!
Oh, and about that research conducted by the CDC in response to Obama’s executive order, here is a snippet (though I wouldn’t call it research exactly, more like a summary):
They identified all homicides in the largest counties in Washington, Ohio and Tennessee that took place in the home of the victim. For each case, they matched a control based on: 1. Living in the same neighborhood as the victim; 2. Being the same gender; 3. Being the same race; and 4. being in the same age bracket as the victim.
You do understand that they couldn’t assign people to the homicide group, right? I’m starting to wonder.
You do understand that they have to ask them questions first to find out things, right? Like their history of illicit drug use or the history of violent behavior in the home, right? You have to enroll people in the study before you can ask them these things.
You’re not going to find a doppleganger for everyone. When you can simply measure important factors and control for them, and you know that multiple factors will influence who gets killed, it’s stupid to suggest that you should fine tune the comparison based on a single factor. When you do that, you are introducing a manipulation that makes the comparison less generalizable and affecting the meaning of the other factors. As you yourself note, there are several. Which do you choose?
More importantly, why? Why do this when you can measure them all and test their relative effects on the outcome?
Just because it seems like sorcery to you is no reason for the rest of us to struggle in ignorance.
And Kellermann is tainted! Tainted by the CONSPIRACY!
I am clearly using odds and ratio interchangably here. The distinction I am making is not the difference between odds and rate. The distinction I am making is clearly about how you are PHRASING things. You phrase things to imply a causal relationship between guns and murders in the home. You were implying that keeping a gun was the source of those increased murders. And yet you focus on the use of terminology like odds versus rate; this is what I mean by pettifoggery.
So what substantive things do you have to say other than taking issue with my loose use of statistical terminology that NOONE else seems to be getting confused by?
Do you understand that none of this shit even IMPLIES causality? My guess is that you initially disliked my use of the term correlation because everyone knows that “correlation /= causation” Of course the same can be said for association and odds ratio or anything else these studies have to say. NONE of them implies causation. So you take issue with my use of the term correlation (a distinction without much of a difference) and then continue on your path of pettifoggery to try to convince readers that you are actually saying something substantive.
You want to be able to say (actually I think you do say) murders double when there are guns in the house ‘it is as simple as that’ But its not that simple, now is it? I wonder how other scientists feel about how you are using “science”
He knows what I mean, its just not convenient for him to acknowledge the possibility that the studies can be interpreted in any other way than to imply some sort of causal link between guns and these murders.
What do they mean by intimate acquaintance (the words seem to contradict one another a little bit, don’t they? Acquaintance implies a distant relationship while intimate implies a close relationship)? We already know that ~30% of murders committed in homes with guns are committed by family members and ~25% of ALL murders are committed by family members so who are these intimate acquaintences that seem to be making up the difference to get to 7.8 times the odds of getting murdered?
Are rival gang members or criminal co-conspirators considered intimate acquiantances? Or are we talking about lovers. I can’t find the stats for murders by intimate acquaintences in the general murder stats but ~55% of all murders are committed by people who knew the victim.
Yeah because the NEJM has never published anything with almost no real probative value.
Once again, you are confusing causation and “correlation”
Perhaps these folks would have died at much higher rates without those guns. These people are not fungible. Sure they control for somethings but I still don’t see where they control for things like gang membership. Or do you think that arrest record is enough to capture all of the factors I am talking about?
Actually, I think you can control for all this stuff… That is mostly what the peer review does. It checks whether they did the math (or science) correctly, which includes for controlling for variables.
What peer review does not do is check whether the data is worth a damn for any non-political purpose. What it does not do is confirm the opinions of the authors of the study.
The problem is that as you point out the victims weren’t merely a population of gun owners. They were a population that had lots of things that might make you think they were riskier. The fact that folks who had a gun and lived alone did not get murdered much more frequently than people who lived alone and DIDN’T have a gun makes you wonder if maybe there might be some sort of common cause underlying both these factors but as far as “some people” are concerned, the guns are the cause of the increase.
For doctors, proving causation is not critical before throwing red flags on the field. If they see a strong enough correlation between cold weather and exploding eyeballs, they might tell everyone to move to warmer climes. Even if the actual cause is some undiscovered variable, the correlation is enough to make them want to act. The hope in some cases is that you might eliminate the unrelated reason by eliminating the correlated variable, maybe HVAC systems cause the exploding eyeballs.
Kellerman is tainted by his bias. It may affect how he interprets the results and how he conducts the study. The notion that this bias may exist in a lot of the people that conducted this research is hardly conspiracy theory.
Damuri, every time I say the studies show that the odds of homicide are doubled for people with a gun in the house, you say something like “no they don’t” or “you’re confusing correlation with causation.”
THE ODDS RATIO DOES NOT MEAN CAUSE. It is a measure of association. It’s like if a study showed (somehow, setting aside the issues of correlation for dichotomous variables) there was a correlation of .60 between guns in the house and homicides.
You’d look pretty foolish if you spouted “you’re confusing correlation with causation” after I told you what the correlation was, right?
The same is true here. I’m simply going to stop replying to anything else from you on this until you demonstrate comprehension of this.
Are you under the impression that I claimed the NRA has conducted research?
You make this claim that they identified all homicides, and that for each they matched a control. This again is your misleading presentation of what the actual data says. In fact, the study itself states this:
However, when you look at the study, there were actually 444 homicides in the home during the study period. Of those, 24 were excluded for various reasons noted in the study, and another 15 had no proxy respondents for the case subjects. After that, 388 matched pairs between case subject and control subjects were obtained through interview (gasp! these surveys must be completely reliable instead of the other surveys that you don’t like - those results make the surveys totally worthless!). 388 / 444 is 87.4%. But when it came to matching more complete data to conduct the multivariate analyses, there were actually only 316 matched pairs. 316 / 444 is 71.2%. Hardly “all”. Whether this is intentionally misleading or not, it provides opportunity for selection bias.
From Table 1, of the 420 homicides identified, only 209 were committed with a firearm. The other 50% were committed by other means. From that, Kellermann feels confident stating,
This gives the impression that the firearm is what presents the danger. It would be much less newsworthy if the conclusion stated that 'after matching for four characteristics and controlling for the effects of five more, we found that the presence of one or more firearms in the home was strongly associated with an increased risk of homicide in the home (adjusted odds ratio, 2.7), 50% of which were committed with non-firearms. It’s like saying that a firearm in the home increases your chances of getting stabbed or burned to death.
From the study:
You’ve already listed the characteristics that were matched, here is what was controlled for:
[ul]
[li]Home rented[/li][li]Living alone[/li][li]Household member hit or hurt in a fight in the home[/li][li]Household member arrested[/li][li]Household member used illicit drugs[/li][/ul]
Here are *some *things that were not controlled for:
[ul]
[li]Drinking caused problems in the household: 24.8% vs. 5.7%[/li][li]Any household member had trouble at work because of drinking: 9% vs. 0.8%[/li][li]Any household member hospitalized because of drinking: 11.4% vs. 2.3%[/li][li]Any family member required medical attention because of a fight in the home: 17.3% vs. 2.1%[/li][/ul]
Yes, there will not be a doppleganger for everyone. Would you agree that the more closely the control subjects match the case subjects, the better? This is the source of my previous question that you evaded. At what point of disparity between the case and control groups would you begin to discount the results? Would a study that paired 3 time violent felons as the case group with only suburban housewives with 3 or more kids be as valuable as one that paired the felons with other felons? Because of the 4 things that were matched by Kellermann, this example would qualify in this study.
Do you find it surprising that nearly all demographic AND behavioral factors in Table 2 and Table 3 indicate either similar or higher levels of riskier characteristics among the case group? Kellermann is quick to attribute the results to the presence of a gun in the home. It seems more likely to me that it is the behaviors and characteristics of the person, rather than the objects they own. Treating the 2.7 figure as if it applies equally to everyone is silly and lazy. I’ve read the mantra many times - Dont go to stupid places, don’t associate with stupid people, don’t do stupid things. That’s going to reduce the chance of homicide greater than whether or not there is a gun present in the home.
I fully understand what you keep saying. I have understood for a while now. If you didn’t use the study to IMPLY THAT THERE IS SOME SORT OF CAUSAL LINK, then I would not be accusing you of conflating correlation and causation. After all I don’t dispute the factual findings of the study. Among the population of people that had guns there was a higher chance that there was a murder in the home.
Instead you say things like:
“people who kept a gun at their residence - were twice as likely to be killed as people who did not. It is that simple.”
You further imply causation by saying that if I forced a million random households to keep a gun in their homes then the studies say that those million households would experience a 240% increase in murders. The studies say no such thing. The studies just tell us what actually occurs among households that (for whatever reason) voluntarily chose to keep a gun in their home.
You say that the increase is largely due to family members killing each other, and yet about the same percentage of murders are committed by family members in the general population as there are among households that have a gun in the home (assuming 25% ~30%).
And you keep saying that the distinction between words like odds ratio and correlation and rate are so fundamentally important in understanding the studies that my misuse of these words makes my understanding of these studies flawed in some way. In what way does my misuse of these words change anything of substance? Or do you just want to remind people that you are a scientist so your interpretation of this fairly simple study is entitled to some more weight than anyone else’s. Either that or you are engaging in pettifogery and semantics