Crimes that warrant the use of deadly force better involve firearms. So the 5x stat most certainly is an isolatable statistic.
For example, if a black youth is 3 times more likely to have gun and 4 times more likely to have a criminal record, then they are 12 (3 X 4) times more likely to have the combination both having a gun and a criminal record.
Ignoring the fact... wait... lets not ignore it, let's cut that arithmetic failure in its place. First the most blatant:
1) 3x4 = 12. Yes, it does. How that is even remotely applicable to the subject, however, is lost on me. It would be 3 + 4, assuming some bad assumptions. That'd be 7x, not 12x.
Combined probabilities of two things occuring are multiplicative of the odds of each thing, not additive. Thus, differential probabilities across factors are also multiplicative.
No it isn't.
1,000,000 whites
50,000 white people have guns (5%)
80,000 white people have criminal records (8%)
130,000 combined assuming no crossover
1,000,000 blacks
150,000 blacks have guns (15%) 3x rate of whites
400,000 blacks have criminal records (40%) 5x rate of whites
550,000 combined assuming no crossover
550,000/130,000 << 21
Is the relative math wrong?
What is wrong is your use of addition and not multiplication. The relevant question is "How many blacks have
both a gun
and a criminal record?". By adding the raw frequencies, you are answering a very different question of "How many blacks have
either a gun
or a criminal record?"
doubtingt, the number I gave you, that is lowest possible value. If you assume there is no overlap, that is the maximum you can possibly see. Based on 3x and 5x for both stats.
AAAHHHHH!!! No it isn't. Why don't you grasp that you have to multiply the rates of each variable to get the probability of being in the "both" group? That is the issue because that is the group most likely to be shot by cops.
Because these aren't probabilities, but flat out statistics.
They are probabilities of having various combinations of factors. What definition of probability are you (mis)using that has neccessary properties absent from the current analysis? It is about co-occurrence probability and is same as computing the probability of tossing two heads in a row. You take the probability of a heads outcome on one toss (.50) and multiply it by the probability of heads on the other toss (.50), and you get .25.
We have 2 yes/no variables that when combined create 4 groups of 'neither', 'Gun but no record", "record but no gun", and "both".
But these aren't random values. There are weights involved here. A person with a gun is much more likely to have criminal record and vice versa. Those with criminal records are more likely to have a gun.
That has zero relevance. It does not change the fact that we have four types of combinations of the two variables, and that blacks and whites are deferentially distributed among those 4 types, as a multiplicative function of their relative rates for each variable. The partial co-dependance of the variables doesn't change the math that determines the racial difference in probable membership in each of the 4 types. It only increases the overall % of all people who are in the "both" category
and reduces the overall % in the "one but not the other" categories.
A cop will be more likely to shoot if a person has either one of those in isolation than neither of those. But the probability of shooting goes up exponentially if the person has both, and even more if they also have other qualities, such as advertised membership in a violent and well armed gang, have recently committed a crime and thus have a strong desire to resist arrest, are in the company of others with one or more of these qualities, etc..
But the statistics you are trying to use to demonstrate that are in no way relatable.
The stats about the probability of having various combinations of "risk factors" is separate from the issue of the psychological causality that prompts cops to shoot.
That psychological causality and the interaction of causes is just based upon trying to be coherent with pretty much every relevant piece of data and validated theory of human fear, decision making, and risk assessment. I am applying it to cops shooting decisions in a very straightforward way that assumes they are human beings and thus the mountain and century of data from the behavioral sciences applies.
What you are saying is that because x times young blacks have guns, they are x times more likely to be shot. The reality is that because x times young blacks have guns, may be y times more likely to be shot. You are going 1 to 1 here, that isn't correct.
You are the one trying to compare the 21 with the 3 and the 5, and solely based upon 21 being larger, you infer those variables cannot explain the gap, so it must be (God) racism. I am just exposing the error in your logic that each factor should be evaluating in isolation and/or that combined factors do not increase the relative risk at all above the single factor with this largest difference. Exactly how much each factor translates into a difference in risk (your X and Y point) is not relevant to your wrong math and failure to consider causal interaction and how differences on each factor multiply across factors to increase the difference above whatever the largest difference is on any single factor alone. You are introducing a new issue with the X times guns leading to Y times more likely shot argument. But it isn't relevant to the other issues. A gun could increase the prob of being shot by 1000 times. The bigger the impact on being shot, the more of the 21X gap that is explained by racial differences on that variable. But the issue is that if a gun increases it by 1000 times, then having a gun plus a record, plus countless other variables on which blacks and whites differ, will increase the prob of being shot much higher than 1000 times, even with some degree of overlap. It also does not change the fact that being in this most prob to be shot group is much much higher for blacks and much higher than the simple difference on any single factor, because the probability of having those factors co-occur is a multiplicative function of the indiv prob of each variable. Thus, even without any racism, meaning that the impact of having that combo of factors is the same for blacks and whites, the large difference in having that combo of factors will make blacks much more likely to be shot, and pointing to the fact that the difference on isolated factors is less than the 21X gap is utterly meaningless and presumes that all shootings occur only due to the non-interactive effects of one and only one variable.
A person who has all of these, and many black youths have all of these, is exponentially more likely to get shot than someone with just one or none.
Exponentially? That word has a meaning you know. I think the only exponential factor would be "shooting at a cop" would exponentially increase you risk of being shot.
So then you are assuming that nearly all shooting are completely justifies and involve a cop being shot?
No. In no way is that a reasonable reading of what I said. Once you go to the "exponential" you are claiming that it is a strong likelihood.
No. That is not what exponential means. Exponential merely refers to a non-linear rate of increase with each added factor, such that in purely relative terms (and absolute liklihood) the prob goes up more when you add another factor than it did when you added a factor to the previous number of factors. For example, merely having a gun might only increase your prob 10 times, because the cop won't ever interact with you. But dealing drugs draws their attention and now they see the gun, so its presence is far more impactful and 100 times the prob of just the gun alone. The fact that going from zero factor to one factor (gun) only lead to a 10 times increase, but going from one factor to two (dealing) lead to a 100 time increase means the function that captures the relationship between number of factors and prob increase requires an exponential term in addition to a simple linear constant effect. Such non-linear effects are often due to the underlying causal relations being interactive in the way I have repeated described in my examples.
But again, exponential effects are not at all needed to support my point, only that the prob with 3 factors present is higher than with any of them alone, thus it is senseless to argue as you have by comparing the 21 X to the 3 or 5 times differences on isolated factors.
Are implying that a person holding a gun on the street, with a known violent record, a block away from a just reported robbery, cursing at and refusing to cooperate (i.e, put down his gun) is not that much more likely to get shot than a person eating an ice cream cone in their front yard on a crime free day?
I haven't complained about this yet, but the whole idea that a person having a criminal record is an unknown quantity. It has not been demonstrated how often this is known prior to the shooting.
The record has impacts in multiple ways. First, even when unknown to the cop it makes the suspect more likely to resist arrest. Second, it is known when cops run plates, check IDs, or just go out of their way to know gang members and those with criminal records on their beats (which they do). In fact, the longer the record, the more likely the cop will know about it. Since blacks have longer records, the cops are more likely to know that a black person has a record than a white person with a record. Not to mention gang membership makes it more likely that the cop knows the person has a record, and this also increases the odds that cops know blacks with records more than whites with records. This is yet another of those interactive effects that your entire argument ignores.
Only if they actually shoot the officer do their odds dramatically go up?
This is the question. This is what is being raised.
That is not the question. That is a silly question because we know that not everyone shot is shot in response to a cop. That tells us definitively that other factors notably increase the odds, and all of behavioral science would suggest that it is an multiplicative interaction among many of the variables I have suggested plus countless others, and the science tells us that blacks and whites differ on many of these factors.
Are whites talked down while blacks are just killed? That is the point of the discussion.
So, for the 4th time I will ask you how this theory can explain the fact that black cops show a much larger bias in shooting blacks over whites than white cops do.
What valid core theories of psychology make you think that a black cop is less likely to be afraid of and more likely to engage in conversation with a white suspect versus black suspect, to a greater degree than even white cops?
(btw, I can't spend more time on this exchange. I don't know how I can better explain the multiplicative differences in combined probility of having a set of risk factors, and how you completely ignore this and interactive causality in your fallacious efforts to claim that since 21 is > 3, and therefore these factors cannot explain much of the gap. )