Baton Rouge, LA multiple officers shot

[madd0ct0r]

Esquire, i have not accused you of being heretical, merely not providing all the evidence. You argue this paper represents high qaulity data, do you argue that all the other papers from the previous decades are NOT high qaulity data?

The paper is downloading to my phone. When i get some time ill look at its literature review section.

[Esquire]

Esquire, i have not accused you of being heretical, merely not providing all the evidence.

Perhaps, but you did forgo engaging in a rational discussion in favor of assuming I was blinded by confirmation bias. If you think the paper's conclusions aren't compelling, I would very much like to hear why so that we can discuss it usefully.

You argue this paper represents high qaulity data, do you argue that all the other papers from the previous decades are NOT high qaulity data?

Obviously I haven't read every paper on the subject and I'm not going to be drawn into making an indefensibly broad claim. I say simply that this paper is a good one, and disagreements between it and others - a fair portion of which, yes, are of either lower quality or higher levels of abstraction - can't simply be ignored because you happen not to agree with the conclusions.

[Alyrium Denryle]

Obviously I haven't read every paper on the subject and I'm not going to be drawn into making an indefensibly broad claim. I say simply that this paper is a good one, and disagreements between it and others - a fair portion of which, yes, are of either lower quality or higher levels of abstraction - can't simply be ignored because you happen not to agree with the conclusions.

Note: I have not gone through it all yet, but this stands out as a major problem.

I actually have large issues with their data sources regarding police shootings. They say themselves that because their data contains only the 1s and not the 0s, they had to use another data source for the 0s (one city, when their 1 column contains multiple cities). Combine that with the fact that they are using the police department's own self reports to get their suspect-behavioral covariates and the data is basically useless. For example, they used Houston's arrest data to construct their dataset, and filtered it by crime to find the offenses likely to correlate with a justified shoot.

attempted capital murder of a public safety officer--No issue there

aggravated assault on a public safety officer--no problem there

resisting arrest--Big problem here because "resisting arrest" charges are often tacked on to other offenses that may not even be dangerous (protest related, non-violent drug offenses etc), and are often the sole charge. Where people are arrested for resisting arrest.

evading arrest--Someone is running away, not on its face likely to result in a justifiable shoot.

interfering in arrest--Dicey.

The core problem is that THIS is used as their proxy for the Non-Shoots. Combine with data on the shooting cases that is as likely as not at least partially fabricated due to police officers self-reporting their own shooting incidents...

And it is not a reliable dataset. Granted, they were working with severely limited resources insofar as data is concerned and they did the best they could, but if I were reviewing this I would have given it the reject stamp--possibly with the suggestion that they just leave out that section and stick to non-lethal force. They tried to perform a more detailed analysis than their data realistically permitted.

[Esquire]

I can't tell you how happy I am to have a mathematically-literate interlocutor - all I've said - or at least, all I've meant to say; nobody's perfect - is that the fact that statistical analysis of a large individual-level subsample suggests that we might want to conduct equally-rigorous analysis of the conclusions drawn from ecological samples which have previously found an opposite conclusion. As the standard of research in the social sciences is quite poor, particularly regarding issues of autocorrelation and heteroskedasticity, this is especially true in the current context.

I explicitly do not deny that there are serious issues with the dataset of the paper in question - indeed, I believe I've mentioned some of them several times - and, similarly, the operationalizations of definitions are open to question. I, personally, think that they're reasonable enough for useful inference, but as mentioned above this is only the most rigorous study of the subject I have yet to encounter; I am absolutely open to contrary evidence. That this is true is certainly a judgment on the social sciences in general, but until conclusive research is conducted - which is what I've been advocating for this whole thread - we oughtn't dismiss the paper in question out of hand. That's my whole thesis.

[Alyrium Denryle]

I can't tell you how happy I am to have a mathematically-literate interlocutor - all I've said - or at least, all I've meant to say; nobody's perfect - is that the fact that statistical analysis of a large individual-level subsample suggests that we might want to conduct equally-rigorous analysis of the conclusions drawn from ecological samples which have previously found an opposite conclusion. As the standard of research in the social sciences is quite poor, particularly regarding issues of autocorrelation and heteroskedasticity, this is especially true in the current context.

I explicitly do not deny that there are serious issues with the dataset of the paper in question - indeed, I believe I've mentioned some of them several times - and, similarly, the operationalizations of definitions are open to question. I, personally, think that they're reasonable enough for useful inference, but as mentioned above this is only the most rigorous study of the subject I have yet to encounter; I am absolutely open to contrary evidence. That this is true is certainly a judgment on the social sciences in general, but until conclusive research is conducted - which is what I've been advocating for this whole thread - we oughtn't dismiss the paper in question out of hand. That's my whole thesis.

You dont get a PhD in quantitative biology without lots of statistics.

I would not so much say that the standard of research in the social sciences is poor, so much as I would say that the datasets available are very limiting.

Some of the best work done on this subject is actually experimental, where police officers have to distinguish between Black Armed, Black Unarmed, White Armed, White Unarmed, in a first person shooter paradigm. They lose large numbers of points for false negatives, less for false positives, and their reactions are timed (both as a variable, and if they take too long, they lose points equivalent to a false negative). Unarmed persons generally have some other item in their hands (like a phone).

Results indicate that police are more likely to shoot unarmed black people than unarmed white people, and more likely to shoot armed black people than armed white people. They also take less time to reach a decision.

http://psp.sagepub.com/content/39/3/291.short

There are both training effects (in that training people reduces bias), priming effects (prime them to think about black criminals and the bias returns), as well as experiential effects (members of certain gang units and the like are more often biased in their shoot decisions).

There is also an effect from fatigue

http://www.tandfonline.com/doi/abs/10.1 ... 6LOxfkrKV4

[Esquire]

You dont get a PhD in quantitative biology without lots of statistics.

Well said, Doctor.

I would not so much say that the standard of research in the social sciences is poor, so much as I would say that the datasets available are very limiting.

Fair enough, and a better phrasing for what I meant to say in the first place - social research is difficult in the extreme, and many investigators sacrifice validity for reliability, if you see what I mean.

Some of the best work done on this subject is actually experimental, where police officers have to distinguish between Black Armed, Black Unarmed, White Armed, White Unarmed, in a first person shooter paradigm. They lose large numbers of points for false negatives, less for false positives, and their reactions are timed (both as a variable, and if they take too long, they lose points equivalent to a false negative). Unarmed persons generally have some other item in their hands (like a phone).

Results indicate that police are more likely to shoot unarmed black people than unarmed white people, and more likely to shoot armed black people than armed white people. They also take less time to reach a decision.

http://psp.sagepub.com/content/39/3/291.short

I envy you what I assume is your institutional access - I can't see the full article without paying for it. I'm absolutely willing to take your word for it, but could you summarize the procedure for me? What population, n=?, any particular sampling frames, etc?

In any case, from the abstract alone, the cited bias is not generally true; police officers were not affected and the only significant results were among novices - who are famously open to suggestion - and among SWAT-analogues, who are (I assume, I'll cheerfully concede to real data) disproportionately likely to encounter minority dangerous felons.

There are both training effects (in that training people reduces bias), priming effects (prime them to think about black criminals and the bias returns), as well as experiential effects (members of certain gang units and the like are more often biased in their shoot decisions).

There is also an effect from fatigue

http://www.tandfonline.com/doi/abs/10.1 ... 6LOxfkrKV4

This, fortunately, is easily solvable via simple policy changes, assuming the abstract is accurate - again, no full article for me at the moment. Better training* and less overwork are indicated in every field, especially those which have a monopoly on government-sanctioned force.

*Here condensing the first three of your suggestions in this quoted segment.

[Alyrium Denryle]

I envy you what I assume is your institutional access - I can't see the full article without paying for it. I'm absolutely willing to take your word for it, but could you summarize the procedure for me?

I actually dont have institutional access at the moment (I graduated last year and the university cut my access last month). I do have other means. That one however for whatever reason the link to the full paper I would normally access is non-functional. I will get that as soon as I can.

In any case, from the abstract alone, the cited bias is not generally true; police officers were not affected and the only significant results were among novices - who are famously open to suggestion - and among SWAT-analogues, who are (I assume, I'll cheerfully concede to real data) disproportionately likely to encounter minority dangerous felons.

This particular study was done in Boulder CO, so here it would be gang units and SWAT etc, but in certain other jurisdictions, you get different encounter profiles even for beat cops that would likely have similar results.

But yeah, there is a cure. Better training and less fatigue. But to get both of those things certain institutional shifts are required. It is possible. Dallas managed it, ironically given the mass shooting there.

[Terralthra]

I teach at a university with subscriptions to most things. If it's of collegial interest, I may be able to get you particular studies. PM.

[Esquire]

I envy you what I assume is your institutional access - I can't see the full article without paying for it. I'm absolutely willing to take your word for it, but could you summarize the procedure for me?

I actually dont have institutional access at the moment (I graduated last year and the university cut my access last month). I do have other means. That one however for whatever reason the link to the full paper I would normally access is non-functional. I will get that as soon as I can.

Oh, my bad - as I said in PM, I take your word for it, I'm just curious about the underlying research.

In any case, from the abstract alone, the cited bias is not generally true; police officers were not affected and the only significant results were among novices - who are famously open to suggestion - and among SWAT-analogues, who are (I assume, I'll cheerfully concede to real data) disproportionately likely to encounter minority dangerous felons.

This particular study was done in Boulder CO, so here it would be gang units and SWAT etc, but in certain other jurisdictions, you get different encounter profiles even for beat cops that would likely have similar results.

But yeah, there is a cure. Better training and less fatigue. But to get both of those things certain institutional shifts are required. It is possible. Dallas managed it, ironically given the mass shooting there.

Extremely ironically, perhaps even up to 'life imitating art' levels. Although - the above quoted opens up a couple of axes for further moderating variable-controlling research, I should think. We've got a lot of very serious problems which we aren't trying hard enough to solve, it's just that the general narrative may or may not be true and is, in any case, unproductive.

EDIT:

I teach at a university with subscriptions to most things. If it's of collegial interest, I may be able to get you particular studies. PM.

Terralthra, I can get it tomorrow; I also work at a university, I just happened to be home for the evening by the time I got around to Alyrium's post. Thanks for the offer, though!

[Terralthra]

I teach at a university with subscriptions to most things. If it's of collegial interest, I may be able to get you particular studies. PM.

Terralthra, I can get it tomorrow; I also work at a university, I just happened to be home for the evening by the time I got around to Alyrium's post. Thanks for the offer, though!

Since the university at which I teach isn't in the stone age, I can retrieve PDFs from subscribed journals at home via the library's web interface.

[Esquire]

I'm sure I could too, I just haven't quite figured out the process yet - it doesn't come up that often.