Firearm Related Studies Vol 2: Investigating the Link Between Gun Possession and Gun Assault

[Bone]

The idea behind this thread is to discuss one specific firearm related study. Often threads about firearms meander a bit but for the purpose of this thread I’d like to focus on one specific study. There are many different studies and I’d be glad to start additional threads on those as well, but focusing on one at a time I think can be more productive.

This thread is about one particular study, Investigating the Link Between Gun Possession and Gun Assault. Its authors are Charles C. Branas, PhD, Therese S. Richmond, PhD, CRNP, Dennis P. Culhane, PhD, Thomas R. Ten Have, PhD, MPH, and Douglas J. Wiebe, PhD

Some excerpts:

Methods. We enrolled 677 case participants that had been shot in an assault and 684 population-based control participants within Philadelphia, PA, from 2003 to 2006. We adjusted odds ratios for confounding variables.
Results. After adjustment, individuals in possession of a gun were 4.46 (P < .05) times more likely to be shot in an assault than those not in possession. Among gun assaults where the victim had at least some chance to resist, this adjusted odds ratio increased to 5.45 (P < .05).

Study Limitations
A number of study limitations deserve discussion. Our control population was more unemployed than the target population of Philadelphians that it was to intended to represent.
…
Certain other variables that may have confounded the association between gun possession and assault were also beyond the scope of our data collection system and, therefore, were not included in our analyses.
…
We also did not account for the potential of reverse causation between gun possession and gun assault.
…
Finally, our results could have been affected by misclassification of gun possession status.

One of the big weaknesses of firearm related studies is that they tend to ignore any potential benefits. This is not one of those studies and is why I picked this one next. It tried to measure the risk or protective value that possession of a gun might create for an individual. This study, similar to the one in Volume 1 used the case-control method. The case control method is not designed to nor will ever likely show causality so that criticism is not very strong.

The key weaknesses of this study:[ol]
[li]Did not control for location – it assumes that the level of risk applied evenly to all locations within the city.[/li][li]The case examples differed from the control examples – a high percent of the case selections were shot outdoors while the controls were indoors. There were also differences in criminal history, and drug use.[/li][li]The method of gathering information for the controls consisted of phone calls asking about gun possession. The study showed that that it would take a very low % of controls to not report possession to render the results statistically insignificant. Many folks would intentionally lie to phone survey questions.[/li][li]Case participants were determined based on written accounts from police, paramedics, and medical examiners to determine if the victim had at least some chance to resist. This baselining for case examples as a result was highly subjective.[/ol][/li]
This study also suffers from the potential of reverse causation – those that are already more prone to risk of violence may choose to arm themselves as a result.

Here is from the conclusion (citations omitted):

On average, guns did not protect those who possessed them from being shot in an assault. Although successful defensive gun uses are possible and do occur each year, the probability of success may be low for civilian gun users in urban areas. Such users should rethink their possession of guns or, at least, understand that regular possession necessitates careful safety countermeasures.

Like the previous study, this one recommend against gun possession, though they add an additional caveat softening the recommendation. Given the weaknesses in the study, this conclusion is not well supported.

Previous thread:
Vol 1

[CarnalK]

Your number 2 weakness is addressed in the discussion section:

Finally, as this was a case–control study, we had the advantage of being able to statistically adjust for numerous confounders of the relationship between gun possession and gun assault. These confounders included important individual-level factors that did not change with time such as having a high-risk occupation, limited education, or an arrest record. Other confounders that we included were situational factors that could have influenced the relationship under study: substance abuse, being outside, having others present, and being in neighborhood surroundings that were impoverished or busy with illicit drug trafficking.

Are you saying their “statistical adjustments” were insufficient?

[Shodan]

I don’t know if I understand.

Case participants with at least some chance to resist were in contrast to those that happened very suddenly, involved substantial distances, had no face-to-face contact, and had physical barriers between victim and shooter (e.g., an otherwise uninvolved victim shot in his living room from a gun fired during a fight down the street).

Does this mean that, for those with no chance to resist, owning a gun correlated with a higher risk of getting shot under those kinds of circumstances?

I can understand a tendency not to get involved in a gun fight if you have no gun, but how would owning a gun make it more likely that you get shot in your living room, or from down the street? Unless I am reading it wrong.

Regards,
Shodan

[CarnalK]

Oh I just realized that the bit I quoted also addressed your first weakness as well.

[Bone]

@CarnalK

Your number 2 weakness is addressed in the discussion section:

Are you saying their “statistical adjustments” were insufficient?

I’m saying their statistical adjustments are unreliable. The case vs. control group were different enough to render the results not meaningful because the nature of how the adjustments were made would allow the results to be pushed in any direction.

For example, Case participants were outdoors 83% of the time. Control participants were outdoors 9% of the time. That’s a large difference, and when measuring the defensive aspect of firearms, very relevant. Yet the study somehow controlled for this, even though controlling for these types of confounding variables requires all sorts of assumptions.

@CarnalK

Oh I just realized that the bit I quoted also addressed your first weakness as well.

No, it doesn’t. The study attempts to adjust for different areas of the city by the demographic representation of the surrounding area. Useful, but not dispositive. Look at the assumption when discussing location:

We assumed that the resident population of Philadelphia risked being shot in an assault at any location and at any time of day or night. This is an acceptable assumption because guns are mobile, potentially concealable items and the bullets they fire can pass through obstacles and travel long distances.

Because a person could be shot at any time they posit, then being at home in your bed was the same as walking on a public street at the same time of day.

[DrDeth]

Here’s the main issue:
*
However, compared with control participants, shooting case participants were significantly more often Hispanic, more frequently working in high-risk occupations1,2, less educated, and had a greater frequency of prior arrest. At the time of shooting, case participants were also significantly more often involved with alcohol and drugs, outdoors, and closer to areas where more Blacks, Hispanics, and unemployed individuals resided. Case participants were also more likely to be located in areas with less income and more illicit drug trafficking (Table 1).*

These factors were not controlled for, and thus are much more likely to be a cuase that owning a gun. If you are a criminal or in a criminal gang, yep, you are more likely to get shot. No duh.

and this "A number of study limitations deserve discussion. Our control population was more unemployed than the target population of Philadelphians that it was to intended to represent. Although we did account for employment status in our regression models and our control population was found to be representative of Philadelphians for 5 other indicators, having a preponderance of unemployment among our control participants may mildly erode our study’s generalizability. "

Basicially this means they got some senior citizens and other homebodies to come in to do the study for a small stipend. Those pople are obviously less likely to get shot, imho.

This response tears apart the paper,

http://austringer.net/wp/index.php/2009/10/05/gun-possession-and-assault-with-a-firearm-risky-stuff-or-not/
I’ve read the full paper, and the logic or math is incompletely described by which they arrived at even their numerical results, much less the further conclusions that they take. Of course, I’ve had only a brief time where I was engaged in epidemiology research, and that was over twenty years ago, so I’ve done a bit of review, too. …As noted above, it seems that there is a fairly simple way to check the model against reality. If that is done and it validates the model, I’d be somewhat surprised, but I’d be satisfied on the methodological issues that a real result had been obtained. But in the absence of either a transparent model permitting replication of results or the independent check I outlined above, my impression of the study is that it is more a means for assumptions to be converted into conclusions than a solid piece of empirical work.

[Hentor_the_Barbarian]

@Bone

I’m saying their statistical adjustments are unreliable.

Reliability in science means consistent performance across other factors, such as over time, across informants, and so forth. Statistics are math, so you’re in essence claiming that math is unreliable.

The case vs. control group were different enough to render the results not meaningful because the nature of how the adjustments were made would allow the results to be pushed in any direction.

For example, Case participants were outdoors 83% of the time. Control participants were outdoors 9% of the time. That’s a large difference, and when measuring the defensive aspect of firearms, very relevant.

This comment comes from ignorance about the subject. Think about what the purpose of matching is. If there are factors that you don’t want to evaluate as explanatory, you match participants on them. This allows you to test other factors to determine how much they explain differences in the outcome. Again, if you think about it, it would make no sense for two matched groups to have no differences whatsoever between them, since there would thus be no way to evaluate how much of the outcome of interest other factors explained (since there were no factors that varied between the matched groups).

Yet the study somehow controlled for this, even though controlling for these types of confounding variables requires all sorts of assumptions.

Confounding has a specific meaning. It doesn’t make sense as you’ve used it here. In terms of controlling for something, it’s not magic or mystery. It’s math. If you want to argue that the measures are insufficient that’s an argument you can make. But controlling for another variable simply means that you assess the amount of variance in the outcome explained by the control variable, and then see if another predictor still explains a meaningful amount of the variance in the outcome.

It would be like wanting to test how much tobacco use is associated with cancer outcomes over and above whatever differences there are between men and women in cancer outcomes. You find out how much gender explains differences in cancer outcomes, and then see if tobacco use explains a significant amount of the variance not explained by gender.

Because a person could be shot at any time they posit, then being at home in your bed was the same as walking on a public street at the same time of day.

No, nobody said anything like you claim here. They assumed that people were at risk across settings, not that the risk across settings was the same.

[Hentor_the_Barbarian]

@DrDeth

Here’s the main issue:
*
However, compared with control participants, shooting case participants were significantly more often Hispanic, more frequently working in high-risk occupations1,2, less educated, and had a greater frequency of prior arrest. At the time of shooting, case participants were also significantly more often involved with alcohol and drugs, outdoors, and closer to areas where more Blacks, Hispanics, and unemployed individuals resided. Case participants were also more likely to be located in areas with less income and more illicit drug trafficking (Table 1).*

These factors were not controlled for, and thus are much more likely to be a cuase that owning a gun.

You’re quoting from the study results, and claiming that they didn’t control for the very things that they are describing as being related to the outcome? Do you understand what you’re saying here?

and this "A number of study limitations deserve discussion. Our control population was more unemployed than the target population of Philadelphians that it was to intended to represent. Although we did account for employment status in our regression models and our control population was found to be representative of Philadelphians for 5 other indicators, having a preponderance of unemployment among our control participants may mildly erode our study’s generalizability. "

Basicially this means they got some senior citizens and other homebodies to come in to do the study for a small stipend. Those pople are obviously less likely to get shot, imho.

Really? Is that what they did? You know that studies actually describe the methods that were used in the study. It’s in a section called “Methods.”

However, if you want to set up a strawman to then knock down, you could do so. It’s just that it dramatically undermines your argument.

This response tears apart the paper,

http://austringer.net/wp/index.php/2009/10/05/gun-possession-and-assault-with-a-firearm-risky-stuff-or-not/
I’ve read the full paper, and the logic or math is incompletely described by which they arrived at even their numerical results, much less the further conclusions that they take. Of course, I’ve had only a brief time where I was engaged in epidemiology research, and that was over twenty years ago, so I’ve done a bit of review, too. …As noted above, it seems that there is a fairly simple way to check the model against reality. If that is done and it validates the model, I’d be somewhat surprised, but I’d be satisfied on the methodological issues that a real result had been obtained. But in the absence of either a transparent model permitting replication of results or the independent check I outlined above, my impression of the study is that it is more a means for assumptions to be converted into conclusions than a solid piece of empirical work.

That response says “I briefly had some kind of unspecified role in epidemiological research 20 years ago. Based on that expertise, I suggest that this model should be tested against reality.” That response is in short a meaningless word salad, and suggests that the person doesn’t understand inferential statistics whatsoever. The results of this paper are statistical characteristics of the specific data set collected. It’s transparent, in that the authors tell you specifically their methods.

[Bone]

@Hentor_the_Barbarian

Reliability in science means consistent performance across other factors, such as over time, across informants, and so forth. Statistics are math, so you’re in essence claiming that math is unreliable.

Your taking the specific and generalizing to the whole. I’m not saying that math is unreliable. I’m saying the results and conclusions drawn from this study are unreliable because of the specific methodology used.

This comment comes from ignorance about the subject. Think about what the purpose of matching is. If there are factors that you don’t want to evaluate as explanatory, you match participants on them. This allows you to test other factors to determine how much they explain differences in the outcome. Again, if you think about it, it would make no sense for two matched groups to have no differences whatsoever between them, since there would thus be no way to evaluate how much of the outcome of interest other factors explained (since there were no factors that varied between the matched groups).

The objective of the study was to determine the possible relationship between being shot in an assault and possession of a gun at the time. Simplistically, to in case control study you’d want to identify people who were assaulted and shot who were in possession of a gun, and those that were assaulted and shot and were not in possession of a gun. If all other criteria for those two groups were the same, that’d be a great data set. But when you look at the data, 83% of the case participants were outdoors, and 9% of the control participants were outdoors. But the study wasn’t trying to evaluate whether being outdoors was explanatory - so according to your reasoning they should have matched participants on them, but they didn’t. 83% vs. 9% is significantly disparate.

Confounding has a specific meaning. It doesn’t make sense as you’ve used it here. In terms of controlling for something, it’s not magic or mystery. It’s math. If you want to argue that the measures are insufficient that’s an argument you can make. But controlling for another variable simply means that you assess the amount of variance in the outcome explained by the control variable, and then see if another predictor still explains a meaningful amount of the variance in the outcome.

It would be like wanting to test how much tobacco use is associated with cancer outcomes over and above whatever differences there are between men and women in cancer outcomes. You find out how much gender explains differences in cancer outcomes, and then see if tobacco use explains a significant amount of the variance not explained by gender.

(my bold)

In what way have I used the term “confounding” that doesn’t make sense? Your example is related to the point I’m making. You say that* ‘you find out how much gender explains differences in cancer outcomes’* as if that’s just a known thing. Perhaps in the field of cancer research there are much more robust data where that is a fair throw away statement that can be easily done. If gender in cancer outcomes is confounding, then knowing the magnitude gender has on the impact in cancer outcomes is valuable in controlling for gender. But this isn’t readily available in the field of firearm studies. Controlling for being outside vs. inside would mean the study authors would find out how much being outside or inside explains the difference in being shot while in possession of a firearm. How do you suppose they did that? Do you think it required any assumptions or is that relationship known with certain clarity?

No, nobody said anything like you claim here. They assumed that people were at risk across settings, not that the risk across settings was the same.

What do you think it means if they did not pair match based on location? The study pair matched based on time of incident, age, gender, and race. They did not pair match based on location so a person who was shot on the street could be matched against a person who was asleep in their bed because, “the resident population of Philadelphia risked being shot in an assault at any location and at any time of day or night.” If a person who was walking on the street possessing a gun at 11PM was matched against a person who was asleep in their bed not possessing a gun at 11PM, and a study finds the person who is walking on the street is more at risk of being shot, do you think it’s reasonable to conclude it’s because of the gun possession?

[Hentor_the_Barbarian]

@Bone

The objective of the study was to determine the possible relationship between being shot in an assault and possession of a gun at the time. Simplistically, to in case control study you’d want to identify people who were assaulted and shot who were in possession of a gun, and those that were assaulted and shot and were not in possession of a gun.

Nope. That fails logically and statistically. How could you estimate the likelihood of being shot if everyone in the study had been shot? The probability of being shot for that sample would be 1.0! If you tried to run that model in Stata, for instance, you would get an error message to tell you “No, the outcome does not vary.”

In your study, the thing that varies would be possessing a gun, so you could test what factors predict carrying a gun among people who have been shot. The authors of this study cannot be faulted for not running a different study altogether.

If all other criteria for those two groups were the same, that’d be a great data set. But when you look at the data, 83% of the case participants were outdoors, and 9% of the control participants were outdoors. But the study wasn’t trying to evaluate whether being outdoors was explanatory - so according to your reasoning they should have matched participants on them, but they didn’t. 83% vs. 9% is significantly disparate.

Where does it say that they weren’t trying to evaluate whether being outdoors was explanatory? Again, the fact that the matched groups differed on some factors is not a flaw. This confusion may flow from the challenge you ran into above, where you wanted to remove all variation from the outcome of interest.

In what way have I used the term “confounding” that doesn’t make sense? Your example is related to the point I’m making. You say that* ‘you find out how much gender explains differences in cancer outcomes’* as if that’s just a known thing. Perhaps in the field of cancer research there are much more robust data where that is a fair throw away statement that can be easily done.

It is a “known” thing, in that it is a characteristic of the data set right in front of you. It’s like saying that a researcher says what the mean of a variable in a data set is “as if that’s a known thing.” The variance is a measured quantity. There is an amount of that variance that is associated with a variable, such as gender.

If gender in cancer outcomes is confounding, then knowing the magnitude gender has on the impact in cancer outcomes is valuable in controlling for gender. But this isn’t readily available in the field of firearm studies.

This doesn’t make sense, and it feels again like stats is some mystery thing. Once you have measured variables, these properties hold regardless of the thing that was measured. In some fields, the measurements themselves have more reliability. But if 1=being shot and 0=not being shot, it’s the same as if 1=cancer and 0=not cancer. It’s a logistic regression, with other variables explaining the variance in the outcome.

Controlling for being outside vs. inside would mean the study authors would find out how much being outside or inside explains the difference in being shot while in possession of a firearm. How do you suppose they did that?

They measured whether the incident occurred inside or outside, and then entered that into a regression model to see how much of the variation in the outcome was associated with the variation in the predictor.

Do you think it required any assumptions or is that relationship known with certain clarity?

The relationship is a specific mathematical property of the outcome and the predictor in this set of data. It requires the same assumptions that hold for any other variables in a logistic regression model.

What do you think it means if they did not pair match based on location? The study pair matched based on time of incident, age, gender, and race. They did not pair match based on location so a person who was shot on the street could be matched against a person who was asleep in their bed because, “the resident population of Philadelphia risked being shot in an assault at any location and at any time of day or night.” If a person who was walking on the street possessing a gun at 11PM was matched against a person who was asleep in their bed not possessing a gun at 11PM, and a study finds the person who is walking on the street is more at risk of being shot, do you think it’s reasonable to conclude it’s because of the gun possession?

If it were true in their data that every instance of being inside was also an instance of not possessing a gun, that would prohibit the statistical model from disentangling the variation in the outcome associated with location from the variation in the outcome associated with gun possession. On the other hand, if there is sufficient variability in location and gun possession, such that you have enough representation of the various permutations involved, it’s not a problem. Since they report that VIF was below 10 for their models, this is an indicator that collinearity was not substantially a problem here.

[Hentor_the_Barbarian]

@Bone

The objective of the study was to determine the possible relationship between being shot in an assault and possession of a gun at the time. Simplistically, to in case control study you’d want to identify people who were assaulted and shot who were in possession of a gun, and those that were assaulted and shot and were not in possession of a gun.

Nope. That fails logically and statistically. How could you estimate the likelihood of being shot if everyone in the study had been shot? The probability of being shot for that sample would be 1.0! If you tried to run that model in Stata, for instance, you would get an error message to tell you “No, the outcome does not vary.”

In your study, the thing that varies would be possessing a gun, so you could test what factors predict carrying a gun among people who have been shot. The authors of this study cannot be faulted for not running a different study altogether.

If all other criteria for those two groups were the same, that’d be a great data set. But when you look at the data, 83% of the case participants were outdoors, and 9% of the control participants were outdoors. But the study wasn’t trying to evaluate whether being outdoors was explanatory - so according to your reasoning they should have matched participants on them, but they didn’t. 83% vs. 9% is significantly disparate.

Where does it say that they weren’t trying to evaluate whether being outdoors was explanatory? Again, the fact that the matched groups differed on some factors is not a flaw. This confusion may flow from the challenge you ran into above, where you wanted to remove all variation from the outcome of interest.

In what way have I used the term “confounding” that doesn’t make sense? Your example is related to the point I’m making. You say that* ‘you find out how much gender explains differences in cancer outcomes’* as if that’s just a known thing. Perhaps in the field of cancer research there are much more robust data where that is a fair throw away statement that can be easily done.

It is a “known” thing, in that it is a characteristic of the data set right in front of you. It’s like saying that a researcher says what the mean of a variable in a data set is “as if that’s a known thing.” The variance is a measured quantity. There is an amount of that variance that is associated with a variable, such as gender.

If gender in cancer outcomes is confounding, then knowing the magnitude gender has on the impact in cancer outcomes is valuable in controlling for gender. But this isn’t readily available in the field of firearm studies.

This doesn’t make sense, and it feels again like stats is some mystery thing. Once you have measured variables, these properties hold regardless of the thing that was measured. In some fields, the measurements themselves have more reliability. But if 1=being shot and 0=not being shot, it’s the same as if 1=cancer and 0=not cancer. It’s a logistic regression, with other variables explaining the variance in the outcome.

Controlling for being outside vs. inside would mean the study authors would find out how much being outside or inside explains the difference in being shot while in possession of a firearm. How do you suppose they did that?

They measured whether the incident occurred inside or outside, and then entered that into a regression model to see how much of the variation in the outcome was associated with the variation in the predictor.

Do you think it required any assumptions or is that relationship known with certain clarity?

The relationship is a specific mathematical property of the outcome and the predictor in this set of data. It requires the same assumptions that hold for any other variables in a logistic regression model.

What do you think it means if they did not pair match based on location? The study pair matched based on time of incident, age, gender, and race. They did not pair match based on location so a person who was shot on the street could be matched against a person who was asleep in their bed because, “the resident population of Philadelphia risked being shot in an assault at any location and at any time of day or night.” If a person who was walking on the street possessing a gun at 11PM was matched against a person who was asleep in their bed not possessing a gun at 11PM, and a study finds the person who is walking on the street is more at risk of being shot, do you think it’s reasonable to conclude it’s because of the gun possession?

If it were true in their data that every instance of being inside was also an instance of not possessing a gun, that would prohibit the statistical model from disentangling the variation in the outcome associated with location from the variation in the outcome associated with gun possession. On the other hand, if there is sufficient variability in location and gun possession, such that you have enough representation of the various permutations involved, it’s not a problem. Since they report that VIF was below 10 for their models, this is an indicator that collinearity was not substantially a problem here.

[CarnalK]

This shit is complicated which is why it’d be a lot more useful to cite a “response that tears apart the paper” from someone working in the field rather than a guy who dabbled briefly 20 years ago.

[Bone]

@Hentor_the_Barbarian

Nope. That fails logically and statistically. How could you estimate the likelihood of being shot if everyone in the study had been shot? The probability of being shot for that sample would be 1.0! If you tried to run that model in Stata, for instance, you would get an error message to tell you “No, the outcome does not vary.”

In your study, the thing that varies would be possessing a gun, so you could test what factors predict carrying a gun among people who have been shot. The authors of this study cannot be faulted for not running a different study altogether.

I misstated the simple example. Everyone in the case examples were shot. The controls were not necessarily shot, and the impact of gun possession was what was being analyzed. But this stems from your comment about matching. Gun possession and being shot were the things being tested in the study, location was not. Based on your comment regarding matching, the study should have matched people based on location if that wasn’t going to be evaluated as explanatory. And based on Table 2 from the study, it doesn’t describe any explanatory nature of being indoors or outdoors, i.e. location. The study concludes that individuals who were in possession of a gun were 4.46 times more likely to be shot in an assault than those not in possession. No mention of location and the associated increased likelihood as a result. Do you think the location had a greater or lesser impact on the associate risk of being shot? You can’t tell from the data because they are not assessing it.

To clarify, do you think the study adequately controls for location such that the disparity in the case vs. controls indoors vs. outdoors is not a weakness in the study?

[ElvisL1ves]

@CarnalK

This shit is complicated which is why it’d be a lot more useful to cite a “response that tears apart the paper” from someone working in the field rather than a guy who dabbled briefly 20 years ago.

I’m sure we’ll read all about that in a future installment in this series, looking for ways to discredit a paper that reaches an opposite conclusion. They must be out there, right?

[Hentor_the_Barbarian]

@Bone

To clarify, do you think the study adequately controls for location such that the disparity in the case vs. controls indoors vs. outdoors is not a weakness in the study?

Yes. You realize that case-control designs are not the only study designs, I am sure. How do you understand other logistic regression models to account for the effects of covariates?

[Hentor_the_Barbarian]

@Bone

The study concludes that individuals who were in possession of a gun were 4.46 times more likely to be shot in an assault than those not in possession. No mention of location and the associated increased likelihood as a result. Do you think the location had a greater or lesser impact on the associate risk of being shot? You can’t tell from the data because they are not assessing it.

Forgot to address thus too. This is just plain false. They do explicitly tell you the difference between groups on location. How else would you know about it if they did not? You can even calculate the unadjusted odds ratio - the odds of the case group being outside was 49 times greater than the control.

They then also tell you explicitly that they adjusted for this in the regression models.

It is interesting that you imagined that researchers would measure a covariate, would report it and note it as significant, and would then not adjust for it! What do you imagine the reviewers and editor would have said?

Also, if this is a good faith debate about this study, and your limitations have been shown to be seriously mistaken, do you now agree that the authors conclusions are correct?

[Bone]

@Hentor_the_Barbarian

Forgot to address thus too. This is just plain false. They do explicitly tell you the difference between groups on location. How else would you know about it if they did not? You can even calculate the unadjusted odds ratio - the odds of the case group being outside was 49 times greater than the control.

My point is, the nature of the adjustment for location relies on assumptions that when stacked together make the conclusions unpersuasive. By your reasoning, if the study matched on location then the results should be the same because they sufficiently controlled for it. I’m saying, no the results wouldn’t be the same.

Also, if this is a good faith debate about this study, and your limitations have been shown to be seriously mistaken, do you now agree that the authors conclusions are correct?

I notice this habit of assuming ignorance on the part of people who hold opposition views from you. It’s tedious. And no, I don’t agree that the authors’ conclusions are correct, not in the slightest. Nothing you’ve offered has been useful to add confidence to a study where the case examples are wildly different than the controls but of course, everything is controlled for.

[Hentor_the_Barbarian]

@Bone

My point is, the nature of the adjustment for location relies on assumptions that when stacked together make the conclusions unpersuasive.

You’re wrong. If you feel otherwise, please explain specifically what “assumptions” the adjustment relies upon and how this makes the results “unpersuasive.”

By your reasoning, if the study matched on location then the results should be the same because they sufficiently controlled for it. I’m saying, no the results wouldn’t be the same.

Please state specifically how or why the results would be meaningfully different.

I notice this habit of assuming ignorance on the part of people who hold opposition views from you. It’s tedious. And no, I don’t agree that the authors’ conclusions are correct, not in the slightest. Nothing you’ve offered has been useful to add confidence to a study where the case examples are wildly different than the controls but of course, everything is controlled for.

Ignorance means “not knowing something.” I observe when a conclusion or assertion is made upon a foundation of a lack of knowledge.

If there’s no chance for your view on a debate topic to change, it seems that creating a Great Debate on that topic is not an act of good faith.

[Tee]

@Bone

Like the previous study, this one recommend against gun possession, though they add an additional caveat softening the recommendation. Given the weaknesses in the study, this conclusion is not well supported.

I do think the conclusions are well supported but they are specific: * On average, guns did not protect those who possessed them from being shot in an assault. Although successful defensive gun uses are possible and do occur each year … the probability of success may be low for civilian gun users in urban areas. Such users should rethink their possession of guns or, at least, understand that regular possession necessitates careful safety countermeasures. Suggestions to the contrary, especially for urban residents who may see gun possession as a surefire defense against a dangerous environment, should be discussed and thoughtfully reconsidered.*

I can’t see extrapolating these conclusions to areas that aren’t urban and sketchy.

[Bone]

@Hentor_the_Barbarian

You’re wrong. If you feel otherwise, please explain specifically what “assumptions” the adjustment relies upon and how this makes the results “unpersuasive.”

In the example above about cancer with gender being a potential confounding variable, you indicated that you* “find out how much gender explains differences in cancer outcomes”*. So for the current study, the authors would need to find out how much location explains the difference in their results. This adjustment relies on extrapolating over small amounts of data such that the disparity in the case vs. control could easily yield different results. And that because of that, having a case population that differs nearly 10x from the controls isn’t a good control population to base conclusions on. The study itself identifies this weakness in other factors that it attempted to controlled for - a 5% swing in some of the factors would render the results statistically nonsignificant. But yet you would hold out that this is just math and the results are obviously correct.

If there’s no chance for your view on a debate topic to change, it seems that creating a Great Debate on that topic is not an act of good faith.

Oh, I’m willing to be persuaded. Unfortunately you haven’t done so because you know, it’s math.