What is the report from the UK?

[Monkey Turned Wolf]

This is why comparisons of high-density to low-density areas have to use either multi-year stats (as PDG suggested) or stats across all similar areas (as was used in the PDF I linked to). This virtually eliminates these erratic effects on small numbers.

Yup, like I said I'm not too interested in this discussion, so not going into the pdf (and multi-year stats would have to be over decades, which comes with it's own set of issues). Just wanted to correct that line of thinking, since it's a common one, and has the issue listed above.

 

[dvcochran]

and all sides have a political agenda behind them so I literally cannot draw any definite conclusion I trust myself, let alone expect others to trust.

You found at least one consistency our countries have I common.

 

[pdg]

I'm not interested in getting involved in this discussion at all, but want to point out that some of your logic here isn't good science. If a town of any number has 1 crime...that is too small a number to scale to any other town. Let's say that one year that town has 1 crime, and you scale it so the 1mil town needs 10, like you said. That part makes sense. Now next year, there are 4 crimes in the town. Objectively, 4 is not that much higher than 1, and 3 more crimes could easily happen from year to year, but now the 1mil town is allowed 40 crimes, rather than 10, to be considered as safe as the other town. Next year, there is no crime, that 1mil is not allowed anything despite having ten times the population. Finally, a serial arsonist comes to town, and commits 25 arsons over the course of a year. Now that bigger town is allowed 250 crimes in the comparison, versus the 10 from 4 years before. That's the issue that comes when you try to scale-there is a lot more variability with smaller numbers.

I never said it was good or even exact from a science perspective, but just like per capita figures it's the best I can come up with for a random internet discussion.

Those reasons are why you can't scale on a yearly basis, because a tiny variation scaled up becomes huge.

But if you increase the sampling time to 5 or 10 years, those spikes (arsonist on tour, etc) become a far smaller problem.

The 1mil town being allowed 250 one year or none the previous year or 4 the year before turns into 254 over a 3 year period.

Of course, it messes up budgets

 

[dvcochran]

Yes, I am trying to hijack the thread. It got a little heavy.

never said it was good or even exact from a science perspective, but just like per capita figures it's the best I can come up with for a random internet discussion.

Those reasons are why you can't scale on a yearly basis, because a tiny variation scaled up becomes huge.

But if you increase the sampling time to 5 or 10 years, those spikes (arsonist on tour, etc) become a far smaller problem.

The 1mil town being allowed 250 one year or none the previous year or 4 the year before turns into 254 over a 3 year period.

Of course, it messes up budgets

This why I love analytics. I could not convince our son control engineering is the second best kind of work in the world next only to farming. But he is really good at analytics. He has written algorithms that are being used by several sports teams and schools to measure player performance and probability. He is down in Georgia at Shorter University and will soon be heading to Fordham University (in PA) to adapt their system to use his tools and will then spend the season working for the Eagles. Damn proud of our son.

Like our discussion on crime, it is not difficult to find gross data and while there is a lot of quality data, it doesn't give you any quantitative data. In other words, you do not get the details in gross data and even if you did you do not know what it means.

 

[pdg]

@dvcochran - maybe just semantics, but I consider my manipulation of the data as closer to statistics than analytics.

I'm not doing much of what I would call analysis...

Using 'my' method, I am reaching conclusions that tally with my personal experiences and the reported experiences of people I know.

It's entirely possible that my results are being somewhat led by this, but I'm happy to accept it anyway.

If others are happy to accept it too (whether I've explained it well enough for them to have the same understanding as me or not) then great - if they don't want to accept it or want to disagree, that's great too. If they don't understand a bit and want clarification that's also great, as is whether they agree or not after that.

I'm not trying to influence policy, or to influence individuals - I'm just sharing for my own fun.

If that happens to make someone else think and reach their own conclusions, then that's better than great

 

[dvcochran]

You touched on one of the biggest drivers that caused the emergence of analytics; being led by the results in one ruleset.
Statistical process control (SPC) is very commonly used in the manufacturing environment where the goal is to keep the process as narrow as possible to avoid deviation. It works great when you are making millions of the same widget, the same way, with the same materials and tools, all day every day. The quickest way to affect the in-control process is to add a human element. We are just way too elegant and unpredictable. By applying analytics you can find the factors that affect process and order them by magnitude so that you can weigh their effect.
Getting the math down can get pretty heavy but it gives a much better picture of how things happen in the real world.

 

[pdg]

Certainly, I understand the process and importance of analytics.

The issue is it takes time and effort, and the more complex the data set the more complex the analysis has to be to make any sense.

I don't have the time or motivation to go that deep, so I'm intentionally restricting myself to very basic statistical manipulation, which is providing results that I'm willing to accept as "good enough" for the task at hand.

 

[jobo]

Here's a quick PDF that shows how much variation there can be: https://ovc.ncjrs.gov/ncvrw2016/content/section-6/PDF/2016NCVRW_6_UrbanRural-508.pdf. You'll see the urban/suburban areas consistently have higher rates, and rural areas consistently have the lowest rates. Given that this is averaged over the entire country, it's unlikely a spike in any given rural area will correct the difference (since it's likely some of the areas were in a spike during the year used).

There's a strong suggestion that the lower incidence of interaction in rural areas (translation: spending less time around people) is a major factor. If you don't see people as often, you're less likely to punch one of them.

this seems to ignore poverty ! high density equals poor people, poor people equally poor education which competed the cycle of producing more poor people.

general the lower the income in an area the higher the crime rate, if a moire affluent suburb has a high crime rate, its because it is geographic close to a poor area. rural area are not by definition geographic close to anything

 

[Gerry Seymour]

this seems to ignore poverty ! high density equals poor people, poor people equally poor education which competed the cycle of producing more poor people.

general the lower the income in an area the higher the crime rate, if a moire affluent suburb has a high crime rate, its because it is geographic close to a poor area. rural area are not by definition geographic close to anything

In some areas, the higher density actually has a higher average income, though probably that's because of outliers at both ends in the high-density areas. The stats don't ignore that - it's assumed that's one of the factors at play.

 

[jobo]

In some areas, the higher density actually has a higher average income, though probably that's because of outliers at both ends in the high-density areas. The stats don't ignore that - it's assumed that's one of the factors at play.

they ignore it totally, there no mention at all of per capita income, therefor any attempt to interpratate the data to mean that it's high density that causes crime is exceeding misleading, if you pack people who earn 6 figure incomes in to high density housing, they still are not going to be hunting in packs taking money of people at knife point, t?herefore any suggestion that it's population densityty rather than relative poverty that the main driver in most crime is ridicules

 

[dvcochran]

As soon as you to argue/factor for or against per capita income, or any other factor such as poverty it is no longer gross data.

 

[jobo]

As soon as you to argue/factor for or against per capita income, or any other factor such as poverty it is no longer gross data.

?? per capita income is GROSS data