His post is factual (with a bit of opinion). I see almost no appeals to emotion in his post.
Don’t worry, I don’t think there is anything dangerous about you. Noone will be enticed to agree with you the way they might be with iiandyiiii. Noone else on this board has people standing up and saying “well, if manson1972 is getting mad at you, then you MUST be an asshole” This isn’t a swipe at you (noone says that about someone that I get mad at either) it’s merely my observation that iiandyiiii has a reputation for being noticably and remarkably affable, patient and pleasant. People mistake this for level-headedness rationality and logic when its really just a symptom of politeness. iiandyiiii is polite and sometimes wrong. He is no more open to ideas he doesn’t like than you are but I bet a lot of posters think he is.
Aw man 
And I soooooo wanted to be a dangerous poster too! Dream killer!
I think its something you’re born with. Perhaps years of training with a zen yogi might get you there but its relative so if iiandyiiii started training with that same zen yogi, he would probably outstrip you fairly quickly. We all have our strengths and weaknesses. I am double jointed and can whistle with perfect pitch.
A professional statistician came into this thread and said that ProPublica’s analysis was actually pretty decent.
That’s what I call dangerous.
Nothing dangerous about it. Facts are not dangerous because they usually get you to the right result.
There are about half a dozen “ifs” in his statement.
One of his bigger assumptions was that the ages were already clustered in the 14-19 category. They were not.
Perhaps he will return and address the fact that some of his assumptions were incorrect. He may end up at the same place or he may not.
Perhaps you’ll also admit that a reasonable person can look at the ProPublica report and honestly think it’s decent data – not conclusive by any means, but reasonable to consider when looking at the entire picture. And that your dismissal of the data was highly overstated.
Or would that be too dangerous for you?
Perhaps I haven’t spelled it out in enough detail.
in 2005, Pro-Publica came up with this highly sensational ratio of cops killing 14-19 year old black males 21 times more frequently than 14-19 year old white males. This received a lot of criticism at the time but at the time it was something that was worth looking into.
THEN several studies were conducted and they concluded that cops do not kill black men at higher rates than white men. That makes the Pro-Publica calculations no more than interesting factoids. Do you understand why a subsequent study that addresses the “issue” that you think the pro-publica article brings up, makes the pro-publica factoid irrelevant?
I would have probably dismissed the data back then for the same reason that all the critics of Pro-Publica dismissed it. I think its cherrypicking information for maximum sensational effect but I would have said that there might be something to it; especially because at the time I thought cops were in fact killing blacks more frequently than whites. My posts on this board at the time are consistent with this view. But after the Fryer study, the already weak evidence it provided evaporated. And with the change in facts, my view changed. Why hasn’t yours?
It was 2014, not 2005.
The data is extremely deficient – most police departments don’t report on who they kill and their demographic data. These studies have to rely on the limited data that’s out there – and a couple of studies that conflict with another study, all based on a few limited sets of data, can’t possibly be relied on to make anything close to a final conclusion on such a challenging issue. It’s all potentially useful data, but it paints, at best, a very incomplete picture. No final conclusion can be made without much, much more data, and accompanying studies of that data.
Sorry, I knew that. I meant to say 2014 (maybe 2015?) BLM wasn’t even a ting in 2005. I don’t know what happened.
I see, so the thousands of cops killing in the database are insufficient to draw any conclusions but you think that we really ought to consider the pro-publia report that cops kill black men 21 times more frequently based on 54 killings?
You started this subthread with the statement that:
There is no dispute about the 3X number. EVERYONE agrees that black men are killed by cops at 3X the rate of white men. Pro-Publica carved out a subset of the data where they found that cops killed black men at 21X the rate of white men. I could use the same technique to say that cops killed old white men infinitely more frequently than old black men. Or that they killed pre-pubscent Asian boys 3 times more frequently than pre-pubscent white boys.
There is no conflict between studies. There is a conflict between Pro-Publica’s calculation based on a small set of cherry-picked numbers and 2 actual peer reviewed studies.
This is not a “teach the controversy” moment. This is a “hey wait a minute, maybe we ought to revisit the assumption that cops are killing black men like people have been telling us” moment.
How open are you to the notion that cops are NOT killing black men at higher rates than white men? I am pretty sure that I am open to the notion that cops are killing black men at higher rates because this is exactly what I believed before these studies came out. But, when the facts changed, I changed my mind. Why didn’t you?
Why were you able to reach the conclusion that cops killed black me more frequently on imperfect information but now need so much more certainty before you will discard the conclusion that cops disproportionately kill blacks based on what you now consider flawed incomplete data?
Did you always consider the data flawed and unreliable or did it suddenly become flawed and unreliable when the pier reviewed studies were published?
This doesn’t seem to dispute anything I’ve posted. All I’m saying is that there is not nearly enough data to make a conclusion on how large the disparity is for police shootings, and this is a serious and troubling issue that should be studied further (including demanding that police departments cooperate in collecting and reporting such data).
OK, As I indicated I hadn’t seen the raw data. If there was more granularity than it is possible that they looked as more than just the subset indicated. Still in order to check the hypothesis of youth killed these cutoffs seem relatively natural, so it possible that they were naturally derived.
OK, I hadn’t noticed that so I withdraw my comment that the numbers could have been could have been combined and just note that there seems to be consistancy with younger shootings as well.
Then there could theoretically been up to 3160 (80 choose 2) different hypotheses that could have been looked at. Although in this case multiplying the p-value by 3160 would have been overly conservative since many of those subsets would contain too few events to have any hope of being significant, and the other hypotheses are very dependent on each other (the 14-19 subroup will give very similar results to the 13-19 subgroup). So one would probably want to bootstrap or run a permutation test to determine the statistical significance and confidence intervals for the odds ratio.
You could point that out, but then I would ask whether you could calculate a lower confidence interval on that frequency and what the p-value would be. If it was something like 2 white deaths and no black deaths I can guarantee its not statistically significant. If it was significant, than I would say that you have a finding that for some reason police are more likely to kill older white men than older black men and write your own paper. That still wouldn’t change the fact that we had significantly more younger blacks killed.
why?
No, I am saying that it is extremely unlikely (to the point of impossibility) that if there was no difference between the shooting prevalence of blacks and whites, that there would by chance be a subset of the data that would have as strong statistical significance for any age subgroup as what was observed in this data in the 14-19 age group.
As for the 74-79 data, yes you can find a larger point estimate, but the confidence interval and p-value will likely not be significant. If the p-value was say 0.35 that would mean that even if there wasn’t a difference in the level of shootings you would get that type of result purely by chance 35% of the time. IF you cherry picked to find that result than additional increases to the p-value might be necessary to account for it.
I can’t access the full paper, and (unusually) the abstract doesn’t provide any hard numbers so I can’t really comment on this. All I will say is that negative results are often more difficult to interpret than positive results, particularly in multivariate analyses. If you add enough extraneous variables any result can be made insignificant simply because the more variables you add to the model the more difficult it is to model all of the different parameters and so the more uncertain you are about the accuracy of the final model.
Statistically you can never prove a negative. There is always the possibility that there was a small effect that your data was insufficient to suss out. What is important to look at in these case is the confidence interval which even though you can’t prove a negative rules out certain levels of positives. For example if the paper reports a confidence interval for the difference between black and white shooting prevalence of (0.8 - 1.2) then that would be good proof that if there is a difference, it is modest. But if the confidence interval is (0.25-30.2) than that result (although still statistically insignificant since the confidence interval contains 1) wouldn’t contradict the results of the proPublica study.
Another great (and, dare I say, dangerous) post, Buck Godot!
Even though the data is not comprehensive for the country as a whole it is comprehensive for the several jurisdictions that reported comprehensive data; and based on PEER REVIEWED studies of that data, cops in those jurisdictions do not kill blacks more frequently than white men. They DO harass and use force against black men more frequently but not lethal force.
This conclusion is FAR better grounded than the notion that you can look at a 1%-2%subset of that data, do a few calculations, then say that you have good arguments on both sides. The evidence on one side is significantly better than the evidence on the other.
And, frankly you didn’t seem to think the data was so insufficient back when it looked like the data seemed to support your preconceived notions. Frankly (as my posts during the time showed) I also thought the data looked pretty damning and I formed an opinion about it and I believe so did you. Now that we have new facts presented to us by the peer reviewed studies show u we were wrong, my opinion changed. Why hasn’t yours?
Did you read the same post I did?
You haven’t showed this. You showed conflicting data. A professional statistician has endorsed the ProPublica report as reasonable and valid. I accept that the reports you offered may also be reasonable and valid based on the statistics they used.
When valid reports have such widely varying conclusions, then it’s reasonable to suspect that there hasn’t been enough data gathered to make anything close to a final conclusion.
So why are you making one? Is it just too dangerous to say that maybe we need more data?
The post I read continued to find no mathematical fault with the ProPublica analysis.
Is that too dangerous for you?
He effectively went from “its unlikely they cheated” to “its STILL possible they didn’t cheat”
I don’t think so, but then I’m dangerous. Maybe you can’t think dangerously enough to see it.
Right but now you’re not so sure, now are you? Admit it, this might very well be an instance of data dredging. When you have data with these sort of relationships regurgitated like this with no analysis behind it (the way the two peer reviewed studies have), when you have a group with an agenda picking out a series of 54 data points out of over 3000 data points to achieve shocking ratios, you still don’t think something fishy might be going on? Like cherrypicking of data?
And the opposite with older shootings. On average the ratio is 3::1. to the extent that there is a higher ratio one subset, there will be a lower one in other subsets. That’s just math.
In english so the folks at home can understand that it wouldn’t take a lot to cherrypick this data.
I thought I was pretty clear. 6 black deaths with zero white deaths.
And now that you understand the data a bit more, do you think it is possible the data was cherrypicked?
Well, I guess I should say that after the two studies, the calculation didn’t need to be made. It was fine to throw numbers around and stab in the dark before we had any real facts.
Yes, there is a difference if you ignore all other variables. Hell you don’t need to cherry pick the data to see statistically significant differences. The confidence interval is even better when you look at ALL the data. Peer reviewed studies saw these differences and they say that the difference is basically illusory. When comparing like to like, there is no difference between blacks and whites.
Do you dismiss those studies in favor of the pro-publica bullshit as well? Or even put the Pro-Publica factoid on the same level as those peer reviewed studies?
And do you really consider what Pro-Publica did to be a “study”?