Thinking it Through

[Jon]

Colin, I notice when you argue that when someone provides evidence and answers that defeat your arguments, you ignore them. They are inconvenient to you. You don't seem interested in the truth, just trying to prove that you are right. When asking for clarity in what you mean, you don't provide any. Why? Because my earlier post proved how wrong you were. You cannot give me one fallacy. I asked again, but nope, you won't give me one. And I think I know why.

Earlier, you seemed to believe I would have argued against allocating most of the resources for white homicide towards white on white crime. Yet that wasn't my position. And I think the reason you don't want to repeat what you think is a logical fallacy is because you lost it, by having a faulty assumption of what you thought I would say. Your premise for my faulty logic was wrong.

I have only used the same statistics as yourself.

You don't know how to use statistics properly. You misused it in your example. I am good with statistics because I am well educated on the topic.

On a side note, I think you would benefit from switching attitude from trying to be right towards trying to find the truth of the situation. If you are found out to be wrong, admit it. I do.

Edit: Yes, I do use statistics to promote my view. But I go to the data first and see where that takes you. To form an ideology and then cherry pick the data is the wrong way to do it.

 

[Jon]

Just to add fuel to the fire, the UCR is notoriously unreliable as it is a voluntary reporting system and not all jurisdictions participate. I believe it is being replaced by NIBRS. The UCR does not include many crime categories.

Nothing wrong with your fuel moke. Anything factual is welcome. Regardless, you still get the big picture and broad approximations of how big the problem is. There is no need for precision when the differences between groups are so big.

 

[isladogs]

@Jon
You keep presuming my opinions without evidence.
You believe you are good with statistics and are condescending about anyone who disagrees with your use of them.

I didn't 'seem to believe' anything. I merely pointed out what I perceive to be flaws in your own arguments.
You are entitled to your opinions, as am I...whatever my views actually are on this issue.

FWIW, I regularly admit my errors. I can't recollect you doing so but I'm pleased to hear that is the case

EDIT

To form an ideology and then cherry pick the data is the wrong way to do it

I agree with that statement completely

 

[Jon]

@isladogs

You believe you are good with statistics and are condescending about anyone who disagrees with your use of them.

You criticise my use of statistics but when I criticise your use of them, it becomes "condescending". Completely hypocritical.

I didn't 'seem to believe' anything. I merely pointed out what I perceive to be flaws in your own arguments.
You are entitled to your opinions, as am I...whatever my views actually are on this issue.

There were no flaws. I showed you why. Your silence when caught out shows you cannot admit your errors. You misunderstood the argument, misused the data, and hide behind silence because if you tried to explain the fallacy, it becomes apparent there is none.

I agree with that statement completely

At least we agree on something for a change.

 

[AccessBlaster]

Just to add fuel to the fire, the UCR is notoriously unreliable as it is a voluntary reporting system and not all jurisdictions participate. I believe it is being replaced by NIBRS. The UCR does not include many crime categories.

UCR cites NIBRS as a resource in their data collection.

 

[moke123]

@AccessBlaster

https://www.fbi.gov/services/cjis/cjis-link/srs-to-nibrs-the-path-to-better-ucr-data

SRS features a hierarchy rule that ranks the crimes in order of severity and accepts the report of only the most severe crime within a criminal incident. For example, if a criminal incident includes a robbery and a murder, the hierarchy rule dictates that a law enforcement agency reports only the murder.

So if someone burns a cross on the front lawn, defaces the garage with swastikas, and murders the family, only the murder gets reported under UCR.

 

[The_Doc_Man]

Although you believe you are good with statistics, in my opinion you use them selectively to promote your own viewpoint/bias.

Just to throw fuel on the fire, and Colin I'm replying to YOUR post only because it contained a phrase where my comment "works better" ...

According to Mark Twain (Samuel L Clemens), "There are lies, damned lies, and statistics."

If anyone here understands statistics, I do. A decent part of my dissertation research required considering the statistical behavior of complex molecules in a solution. I had to take multiple college courses in statistics. I understand the foibles of poor statistical set-up for evaluating one kind of experiment or another.

Jon's relative treatments that lead to a conclusion of 4 times the probability for black issues than white issues is valid mathematically. Colin's complaint of melding two different sample sets for comparison has merit as well. Paul's injection of a "causative study" however is on stronger grounds from the fact that it looks beneath the statistics to find reason.

Just remember, if I take a shotgun to a firing range and put up some (very large) graph paper to use as a target, I can enter the coordinates of each pellet hit on that graph paper. If I then run a linear regression on the shot pattern, I will get back a straight line. But if I did that experiment, what would it mean?

Isolated statistics can only go so far.

 

[Jon]

Doc, I know you have a good understanding of statistics from previous discussions on The Mind Tavern.

Colin's complaint of melding two different sample sets for comparison has merit as well.

Its not ideal, as I pointed out in my original post. But it is the next best thing. See my comments below that already covered this:

I cannot find the data for 2019, so the next best is from 2016. Out of 2,870 black homicides, 2,570 were black offenders. Source: https://ucr.fbi.gov/crime-in-the-u....016/tables/expanded-homicide-data-table-3.xls

So lets break that down. I will use 2016 and 2019 figures as though they are the same year, just to get an idea.

I would argue that the point has little merit though. Let me provide the evidence of why.

The concept of needless precision comes to mind. When estimating cosmological distances, you could be out 10's of thousands of lightyears. That is a huge distance. But within the scheme of things, more precision is not necessary. You still see the broad picture and make meaningful conclusions.

Let me give the evidence for this. See the graph below, which shows the number of people shot dead by police from 2017 to 2020.

The low and high points in the data for each full year is from 223 to 235. There is roughly a 5% difference between the top and bottom figure. This is why using adjacent years still gives an accurate picture of the big issues, because of regression to the mean when using larger numbers.

So just for kicks, let's see how much of a difference there is if we use each full year against the 2016 total homicide figure.

Police homicide on blacks
2017 = 223/2870 = 7.77%
2018 = 209/2870 = 7.28%
2019 = 235/2870 = 8.19%

It is clear that this is a case of needless precision. The figures are very similar, year to year, with a range of 7.28% to 8.19% over the last 3 years. [Doc, you can argue the denominator figure will change too, and that this is likely to make the variance higher. But we are likely to only be talking a few percentage points here and there. It becomes a mute point.]

For this reason, the argument that you cannot use adjacent years has little merit. The numbers back up that perspective and you can still make meaningful conclusions. Remember that I said "just to get an idea".

 

[The_Doc_Man]

As I said, Colin's object has merit but I also agree that in context it is permissible to do the best you can with what you've got. For a formal study you would be out of bounds instantly. It is OK for a discussion in the current context, where nobody in the outside world pays attention to a bunch of intellectuals, pseudo-intellectuals, and schmucks like us. (I will NOT assign individuals to categories, however, since that would be based on individual perception.)

In fact, I have looked at similar issues and came up with a 3-fold black/white disparity using different data sources. I believe in the context where I did that, I made it clear that it was prima facie evidence but didn't delve into root causes. Using my dad's old fireman sayings, "Just because you see smoke doesn't mean that you know what is burning." (And since he WAS a smoke-eater for at least 10 years in the East Bank Fire Department in Jefferson Parish, Louisiana, he had a chance to pick up a lot of sayings.)

 

[Jon]

For a formal study you would be out of bounds instantly.

I understand your point Doc, but as perhaps you have come to expect from me, I formally disagree.

Let us take life expectancy. Suppose we have data from the last 10 years to show the average for each year, but one data point is missing.

Remove one years data on any point on that graph, and I will give you a very indicative estimate for that missing year. This of course presumes there was not nuclear war during that missing year, which might in fact be the reason the data was missing for that year.

The thing is, much of the data used in formal studies are merely estimates anyway. They are snapshots in time based on imperfect data. When doing medical trials, as you know, confidence intervals are used to estimate the probability of the drug treatment being effective. You can also use confidence intervals to estimate data points you don't have.

One of the great inventions in probability theory, from a practical perspective, is Bayesian inference theory. In the past, when split testing ads, you try to get 95% confidence to prove the null hypothesis. The great thing about Bayesian inference theory is you can still make decisions based on lower numbers. This has practical implications, because when doing ads for example, you often don't get a big enough sample size to prove the null hypothesis. But using Bayesian inference, you can be, for example, 75% confidence that ad A > ad B. i.e. you can make a pragmatic decision that 3 out of 4 times you will have made the right choice.

Bayesian calculator: https://abtestguide.com/bayesian/

As for the root causes, and correlation/causation, that is an altogether different matter. I frequently use root cause diagramming to solve persistent issues and to get clarity. The software I like is called Flying Logic. Here is an example diagram:

Maybe I am just lazy and don't want to face the world!

 

[The_Doc_Man]

My only real complaint, here, Jon, is that to meaningfully use Bayesian methods, you need a decent model of that about which you make a prediction.

Just as an illustrative point, take my shotgun experiment. Let me plot the pellet impacts - except that I missed one. Now let me use Bayesian methods to predict where that pellet impact mark will be. But the big "gotcha" in all of this is that my distribution model that predicts the best straight line is going to give me an answer even if there IS no straight line - because it's the wrong model.

I'm not going to criticize the concept but my question will always be about the model used to make the estimates that are part of the Bayesian method. If we don't know the underlying causes of the biased behavior, we have an incomplete model. Not a dig against anybody, because when it comes to real-world human behavior, every model I've ever seen is a hunk of junk to varying degrees.

 

[Jon]

But the big "gotcha" in all of this is that my distribution model that predicts the best straight line is going to give me an answer even if there IS no straight line - because it's the wrong model.

I completely agree about using the right model. There will also be tons of models that I have no concept of. I just work with the few common ones that I know of.

You want to find the signal in the noise. And you want to try and find out if the signal is a meaningful signal, rather than something that lends no useful purpose. For your gun target example, I would imagine a frequency distribution model that plots the distance from the centre of the target might be useful, since it could give you an idea of how tight or spread out your average shot is. But as you point out, a straight line has no real meaning with the data.

I am curious Doc, did you have a red pill moment after learning your statistics as a matter of necessity?

 

[Pat Hartman]

But the big "gotcha" in all of this is that my distribution model that predicts the best straight line is going to give me an answer even if there IS no straight line - because it's the wrong model.

Sounds like what happened to the COVID models that scared the crap our of the world and caused the massive, universal shutdown for no reason. Once the numbers are in, we'll see that places that didn't use Draconian measures had similar or better death rates than places that did.

 

[The_Doc_Man]

I'm a chemist. Who do you think MAKES those red and blue pills?

 

[Steve R.]

Once again we have an example where the mob and much of the media reacted before sufficient facts were disclosed. However, that does not justify the use of excessive police force. When the police use excessive force, that can't be defended, Nevertheless, those on the left have been inciting immediate violence based on a false narrative, which itself (both the violance and narrative) cannot be defended. Furthermore, these incidents are being forcibly portrayed as racist incidents despite the lack of evidence. This is dishonest.

Medical examiner: George Floyd had 'fatal level of fentanyl' in his system, but is 'not saying this killed him'

A memorandum filed by the Hennepin County Attorney’s office on June 1 indicated that chief medical examiner Dr. Andrew Baker, who listed Floyd’s death as a homicide, thought the amount of fentanyl in Floyd’s blood was “pretty high” and could be “a fatal level of fentanyl under normal circumstances.”

“[Dr. Baker] said that if Mr. Floyd had been found dead in his home (or anywhere else) and there were no other contributing factors he would conclude that it was an overdose death,” the memo said.

As another similar incident, were the mob immediately reacted before all the facts were known. Tucker Carlson says people blaming the police officer for repeatedly shooting Jacob Blake are attempting to “inflame fear and race hate”

News accounts say Blake had an extensive history of run-ins with the law. That would include an arrest for threatening people with a gun and committing violence against police officers. Yesterday, Blake refused police orders to surrender. When he appeared to reach for something in the darkness, police officers shot him. Here is the 10-second video of the exchange now circulating on social media.

 

[moke123]

In September 2015, the Racine County Eye reported that Blake had been arrested after brandishing a gun at a bar in Racine, and that officers required the assistance of a police dog when Blake allegedly refused to comply with their orders. He was charged with several offenses, including: felony resisting arrest “causing a soft tissue injury to a police officer”; carrying a firearm while intoxicated; and endangering safety by the use of a dangerous weapon.

However, Blake was never tried or convicted on those charges, and in August 2020, the Racine County Eye reported that the charges appear to have been dismissed at the urging of the prosecutor and no longer appear in public court records.

“His death was caused by the police subdual and restraint in the setting of severe hypertensive atherosclerotic cardiovascular disease, and methamphetamine and fentanyl intoxication,” officials from the Office of the Armed Forces Medical Examiner wrote. “The subdual and restraint had elements of positional and mechanical asphyxiation. … We concur with the reported manner of death of homicide.”

 

[Steve R.]

brandishing a gun at a bar in Racine

Just more ambiguity to this whole discussion. Was Blake's arrest an unjustified arrest by police, or was the prosecutor a Democratic hack promoting "social justice" by releasing Blake?

Back on June 29th concerning a white couple (Patricia and Mark McCloskey) arrested for "brandishing" guns, you wrote: "I'm not okay with what they did. ... I have zero respect for irresponsible gun owners." Whether Blake's "brandishing" was equivalent to that of the McCloskeys' is probably a debatable unknown.

 

[Isaac]

At first the media narrative on Blake was all about "he was breaking up a domestic violence fight". That morphed into the now-known, "he was actually the subject of the 911 call". We know he told his children, let's go, we're fixing to go, but then went to the driver's side instead and appeared to reach for a knife--all while refusing to obey the police commands, which presumably included something along the lines of turning around or putting his hands up. I'm waiting for all the facts--more facts, more clarification, more information, before coming to judgment on this one.

That information Floyd case is a reminder of something I've said before. I'm not sure it changes the final determination of wrong on the part of the police officers, I'm saying outside/separate from the Floyd case specifically, it does bring back to our minds the reality that police officers are faced with the challenge of not only deciding how much force to use, but also thinking about how much force is OK given that the subject is high on drugs and their heart is on its last legs in the first place. "Eggshell plaintiff"..

 

[moke123]

Just more ambiguity to this whole discussion. Was Blake's arrest an unjustified arrest by police, or was the prosecutor a Democratic hack promoting "social justice" by releasing Blake?

Back on June 29th concerning a white couple (Patricia and Mark McCloskey) arrested for "brandishing" guns, you wrote: "I'm not okay with what they did. ... I have zero respect for irresponsible gun owners." Whether Blake's "brandishing" was equivalent to that of the McCloskeys' is probably a debatable unknown.

No, two wrongs dont make a right.

or was the prosecutor a Democratic hack promoting "social justice"

A native of Racine County, Tricia Hanson has worked for the Racine County District Attorney’s Office for as long as she has been a lawyer.
This week, after nearly 22 years with the office — more than four of them as the deputy district attorney — the Marquette University Law School alumna and Wind Lake resident will step into her post as Racine County district attorney.
Hanson, a Republican, won election to the seat in November after beating Tom Binger, a local Democrat and assistant district attorney for Kenosha County. Incumbent Racine County District Attorney Rich Chiapete did not seek re-election. Both Chiapete and Hanson are Republicans.

 

[Isaac]

the most recent charges brought against Blake were in Kenosha county.