Hello /pol/.
I got tired of seeing statistics be misused on /pol/ and decided to conduct a proper analysis of crime vs. race in the US. This is a serious post and was a lot of work, so please give serious replies and comments. Also, there will be several posts due to word limits, so wait until I finish the last post to reply. The data I'm using is extracted from the US Census Bureau website for population and the FBI website from crime data. Both refer to 2014.
Population data: http://factfinder.census.gov/faces/tableservices/jsf/pages/productview.xhtml?src=bkmk
Crime data: https://www.fbi.gov/about-us/cjis/ucr/crime-in-the-u.s/2014/crime-in-the-u.s.-2014/tables/table-5
Some notes on linear regression before I begin. Linear regression is a statistical technique used to quantify correlation between variables. A dependent variable (Y) is related to an independent variable (X) via an equation of the form Y=A*X+B. The slope A tells us, on average, how much Y there is per unit of X and the intercept B tells us, on average, how much Y there is in the absence of X. The coefficient of determination R^2 tells how good the correlation is. For example, if R^2=0.23, than 23% of the variation in Y is explained by the variation in X.
Pic related is the usual type of picture I see thrown around /pol/, except normally the regression equation is not displayed. The graphs are percentage of black population vs. number of crimes per thousand inhabitants, for all US states. Three immediate observations are (continues in the following post):
>>56799483
>>56799483
1. There is no intercept B, which is the same as assuming that there is no crime in absence of blacks (which of course is not true).
2. A statistical outlier is included in the analysis (which is the isolated point in the upper right of the graphs). In statistics, outliers are data points which deviate a lot from the data cloud (the group of points in the lower left of the graphs). Turns out this outlier is the District of Columbia, which is not state and is more comparable to a city. Since this is an analysis of state-wide statistics, D.C. should not be included.
3. The coefficient of determination R^2 is negative. This means, without going into the technical details, that a linear model is a worst interpretation of the data than simply staring at the cloud of points. This is mostly due to the inclusion of the D.C. outlier.
As a conclusion, the charts you see, which are commonly thrown around /pol/, are absolute garbage.
In the following post I will conduct a proper analysis of this data set.
>>56799558
Excluding the D.C. outlier, which is not a state and therefore has no place in a state-wide evaluation of crime rates, and including an intercept B in the regression equation, the result is pic related.
The initial conclusion is that, without considering any other factors, the variation in the percentage of black population explain 14.2% of the variation in the rate of violent crimes and 12.9% of the variation in the rate of property crimes (see the values of R^2).
Another major concept in statistics is the level confidence in your results, and I don't mean this is a subjective way. The slope A and the intercept B relate the average trends between X and Y, but they do not account for the dispersion in the data. This can be accessed by calculating the standard error associated with A and B. Again, I won't go into the technical details, but you can read about linear regression on wikipedia if you'd like. For this particular example,
Violent crime: y=(0.050±0.018)*x + (2.933±0.255)
Property crime: y=(0.215±0.080)*x + (23.234±1.150)
(continues)
>>56799483
>I got tired of seeing statistics be misused on /pol/
Boo hoo
>>56800100
To ensure that there is in fact a correlation between Y and X, the error range of the slope A must not contain zero. Using the standard error in the previous post, we can evaluate the error range of A with a confidence of about 68%.
Violent crime: A range =[0.032; 0.068]
Property crime: A range =[0.135; 0.295]
So we can say with 68% confidence that crimes rates and percentage of black population are related. For higher confidence, we need to multiply the standard error. I increased the confidence level until the lower end of error range in A until was just touching zero. The conclusion is that the relation bet crime and black population holds with 99.5% confidence for violent crime and 99.25% confidence for property crime.
So, I can say I am fully confident in this particular analysis. My final conclusion is that, without looking at any other factors, the percentage of black population explain 14.2% of the violent crime rate and 12.9% of the property crime rate.
I the next post (the last) I will write a few extra notes about other factors and how they may alter the conclusion.
>>56800644
True as this may be, your numbers are being thrown off- Illinois is practically 2 different states, chicongo and the rest of it, and the low crime countryside offsets the very real difference. I would very much like to see a comparison of black cities very white cities.
![IS1Nejx[1].png](https://i.imgur.com/hFwTFMG.png "IS1Nejx[1].png")
>>56800808
>I would very much like to see a comparison of black cities very white cities.
Tldr?
>>56800644
When multiple factors affect a single dependent variable, the appropriate procedure is multiple linear regression. For example, if you have Y which is affected by X1 and X2, you fit a model of the type Y=A1*X1+A2*X2+B. However, it may happen that X1 and X2 are correlated. For example, I'm currently looking at the median income per capita across the states, and there is a correlation between crime and income, and also between income and racial composition. So, both race and income contribute to crime, but race and income are also related amongst themselves. There are methods to evaluate this (cross-correlation), so its important to state for now that the conclusions I've presented so far are only valid with looking at any other variables besides crime and race.
If /pol/ wants I can do a multiple regression with crime, race and income data. Its a lot of work though, so I will stop for now.
>>56801100
Ultimately, the most telling comparison will likely be the same one you did, but with cities as opposed to states.
>>56799483
that statistical regression line in your picture is fucked up, there's no reason for it to be pointed so high up when it's basically the same all the way through except for that last outlier point
Either you haven't plotted all your points, or that line is straight up fabricated, there's no reason one point should influence the line that much
also, it's generally accepted to include DC in state comparisons on crime
pic related, much better graphic
I prefer tables to graphs any day
and here's the kicker
>>56800808
>>56801301
In statistics, the larger the sample, the more reliable the results. Scaling-down from the state level to the city level decreases the sample size.
>>56801033
This shows qualitatively that there is cross-correlation between crime, race and income. Do you have a multiple regression analysis based on this dataset?
>>56801388
My first post serves to debunk the charts and show how and why they are the result of improper analysis. The following posts discuss how to do things correctly.
>>56801072
If you are ignorant of statistics, anyone can fool you into believing anything. Thats the main tl;dr. For conclusions on the example of crime vs. race, read on.
>>56799483
>I got tired of seeing statistics be misused on /pol/ and decided to conduct a proper analysis of crime vs. race in the US.
You have to check for single moms.
Once you put single moms into the picture, it explains everything. Nigs are nigs because their higher rate of single motherhood. Nigs raised by mom+dad behave normally. Not nigness causes the problems, but cunts.
Free roaming cunts are the cancer of society.
https://randomcriticalanalysis.wordpress.com/2015/10/31/on-philip-cohens-knee-jerk-response-to-chettys-causal-mobility-data-and-its-association-with-single-motherhood/
>>56800339
>>56801428
What you guys need to understand is that statements such as "half", "double" or "triple" have no meaning in statistics unless you can calculate the dispersion relative to the average and the level on confidence in your analysis.
Stat friends is there a way to do a multi-national analysis of niggers on a global scale? Like compare %nig/crime for Sweden, China, Japan, etc to show a nog trend
>>56802159
implying there are enough chinese nigs for a decent sample
>>56802100
Look at your red data cloud for example, its heavily skewed towards the right of the graph. Why is a linear trend used for a data distribution that is not linear/normal? The red data cloud is very dense towards the right side and very sparse towards the left, with sparse tails you should be treating it as a Poisson distribution.
>>56802159
In western Europe there are records of crime on a racial basis I think. Basically, racial profiling is prohibited by most constitutions in Europe I think. Though, generally speaking, violent crime rates are way lower in Europe and black minorities are much better integrated into society. In Europe the racial disparity is more obvious when looking at middle east/north african immigrants.
>>56802566
Read the source before complaining here, I gave you the link.
>>56802566
i meant to say no records
TL;DR version:
Blacks commit more crime.
>>56802145
Fellow statsfag here. This guy gets it.
Amazing how many people will immediately see a display of data and make assumptions without knowing skite about stats. Dues vult.
>>56802841
btw, It's good work OP but it's too dry for a meme board
>>56802642
I don't need to read the source to tell you that the chart you've presented is bullshit (as in improper statistics). Not saying the conclusions are wrong, but that chart is wrong. Will look at the source later.
>>56802841
No. Here's the proper conclusion: Ignoring other potential determinants of crime, 14.2% of the variation in the rate of violent crime and 12.9% of the variation in the rate of property crime is explained by the variation in the percentage of black population, in the US.
Emphasis on "without considering any other factors" and "in the US". I can't extrapolate outside my data.
The TL
>>56801809
>In statistics, the larger the sample, the more reliable the results. Scaling-down from the state level to the city level decreases the sample size.
Each state has at least 1 capital, regress the crime rates to % of blacks in the ten largest cities for each state and you will get 500+ data points.
Your R^2, slope coefficient and confidence level will all increase.
I would love to be proved wrong
Thank you for your earnest work. It is strange that you need such a complex proof for something as simple as saying "the sky is blue"
>>56803236
>No. Here's the proper conclusion: Ignoring other potential determinants of crime, 14.2% of the variation in the rate of violent crime and 12.9% of the variation in the rate of property crime is explained by the variation in the percentage of black population, in the US.
Only when you use that small sample (shit quality data set)
>In statistics, outliers are data points which deviate a lot from the data cloud (the group of points in the lower left of the graphs). Turns out this outlier is the District of Columbia, which is not state and is more comparable to a city
Yeah, convenient that one of the blackest parts of the US becomes an outlier because "its more of a city than a state". Great reasoning.
>I got tired of seeing statistics be misused on /pol/
So why continue the trend?
>>56801447
Lol at the .82 incidence of DUI.
>>56803523
>>56803893
My sample has a total population of 308 745 538 individuals, of which 40 250 635 self-identify as black. The total of violent crimes is 1 197 987 and the total property crimes are 8 277 829. So my sample as large as it can get without going outside the US. What you'd like me to do is cherry-pick my data. No thank you.
Also, the main between cities and rural areas is the population density. This is unrelated to race. See >>56801100 for how to account for multiple factors.
>>56803662
You are in disagreement with the poster before you. He thinks my results show a rather low correlation between crime and race. I guess subjective judgement is in the eye of the beholder. Also see >>56803236
What you're mentioning
>>56799483
>This is a serious post
How to tell when a post is bait
>>56803893
Also, the District of Columbia is not a state, no matter how much you'd love it to be for the purposes of this analysis. Why not mix state-wide data with a set of hand picked cities so results will please you? If I'm to look at statistical trends on a state-by-state basis, I use state wide data. No argument there.
>>56804118
This is a serious post. Rather series of posts. The first and second entries show how the OP pic is bullshit, and then proper analysis follows.
>>56804284
>If I'm to look at statistical trends on a state-by-state basis
whos cherry picking now
>>56804613
Explain how looking at the total US population is cherry-picking, organized the largest possible denominator without going outside the US is cherry-picking.
I think I will repeat this: My sample has a total population of 308 745 538 individuals, of which 40 250 635 self-identify as black. The total of violent crimes is 1 197 987 and the total property crimes are 8 277 829. So my sample as large as it can get without going outside the US. What you'd like me to do is cherry-pick my data. No thank you.
>>56799483
>This is a serious post and was a lot of work, so please give serious replies and comments.
stopped reading, started laughing here
>>56804997
Good for you friend. Ignorance is bliss.
Oh really.
>>56805131
I used data from 2014.
Violent crime rates per state where obtained from Table 5 of the FBIs crime survey.
US black population estimates where from the Census.
Seems to me that you're spouting shit OP.
>>56805363
Can you show the actual numbers you used? A screenshot is fine, just indicate which column is X and which is Y.
>>56799483
You should also check for metropolises. Crime concentrates in metropolises and it could be that there are disproportionately many nignogs in metropolises.
>>56805363
Also, I think you included the District of Columbia, which for the reasons I've stated is an outlier.
>>56805600
If locally there is a higher concentration of blacks and higher density of related crimes, this doesn't change the number of crimes per black. This the important figure, crimes per capita. You experience more crime in large cities or even specific neighbourhoods due to high local population densities.
>>56805707
>Also, I think you included the District of Columbia, which for the reasons I've stated is an outlier.
just stop posting
>>56805858
I'm going to cite myself again:
lso, the District of Columbia is not a state, no matter how much you'd love it to be for the purposes of this analysis. Why not mix state-wide data with a set of hand picked cities so results will please you? If I'm to look at statistical trends on a state-by-state basis, I use state wide data. No argument there.
I think I will repeat this: My sample has a total population of 308 745 538 individuals, of which 40 250 635 self-identify as black. The total of violent crimes is 1 197 987 and the total property crimes are 8 277 829. So my sample as large as it can get without going outside the US. What you'd like me to do is cherry-pick my data. No thank you.
>>56805858
Also, show me the base numbers, just take a screenshot of the data or post a plot.
>>56804776
>Explain how looking at the total US population is cherry-picking
Because you are purposefully understating the dispersion in your regressor
OP, Here's my 2 cents on the matter.
Black people are poor and violent due to their own culture and lack of family values. They glorify crime and violence, while hating the police and any organization of law. They have premarital sex without protection, and the fathers always end up leaving. This rises the number of single mothers, and because of that, the kids end up being raised like shit and turn to gangs for that "family" that they lack. The lack of education isn't the problem, in Washington D.C. the spending on education is one of the highest in the nation, yet D.C. is the murder capital. It isn't whitey keeping them down, it is their OWN DAMN FAULT.
Also
>Serious replies
>/pol/
kek
>>56805842
Yes, it does. It changes everything. If people are more criminal in larger cities, and nearly all blacks live in large cities (while whites live all over the place) then the higher black crime rate can be explained by the size of the city.
This is how it's explained why countries in Europe have less murders per capita, because they have far less larger cities.
I know that "more blacks = more crime" sounds good for stormfags, but if you want to find really the root of the black problem and not just spew stormfag memes then you should analyze the influence of metropolises and of single moms. Because then the solution to black crime is completely different.
>>56806337
Its important to know the difference between relative and absolute numbers.
>>56799483
It would be nice to see a multivariate analysis of this, because there's more to it than just percentage black. I bet you could get a correlation of 0.8-0.9 by including other demographic factors. Throw in firearm ownership rates, and leave them out to watch the results not change if you want to watch peoples' heads explode.
>>56806337
Can we just start calling them stormniggers?
Why do blacks rape more than any other per capita? Becuase they live in cities?
Honestly the only solution that will guarantee reduction of crime is to have
mandatory penis inspection in public schools and prisons
>all male post-puberty with cocks 6" or larger are required chemical castration or sterilization during a public school examination
>all repeated offenders in prison will have their testicles surgically removed by animal veterinarians and required to take estrogen pills
No more black single moms
No more high testosterone aggression
Feral animal control of violent blacks
Low cost implementation and efficient if done in public schools and prisons
>>56806017
The data is supposed to be from 2010.
The citation below the pearson coefficient is from 1996.
Also, the Pearson coefficient is the square-root of the determination coefficient, so the R^2 for this trend would be R^2=0.76989^2=0.59.
>>56806141
So you are saying that in order to evaluate the average correlation everywhere between crime and race, I should look at the specific locations with the highest concentration of black population.
This is wrong, if I did this my data range would lack information of regions with low black-population, and would thus be skewed. The appropriate method is to look at crime per capita across the largest range possible of black density, so that I can access the importance of black density in crime. I'm assuming you didn't miss this, but I'm using crimes per thousand populations.
>>56806183
Your post is bullshit for no other reason that you are talking out of your ass. Data? Trends? Something???
>>56806337
If I had to speculate, probably there are a lot less murders per capita in Europe because the poverty rates are lower (the income per capita is higher in the US, but the disparity is lower in Europe, so the poorer in europe are less poor) (...and maybe gun regulations...).
>>56806589
See >>56801100. Saying stuff like "I bet you could show a higher correlation" without actually doing the work is stereotypical racism.
>>56807221
But there are more whites in poverty than blacks as a whole pop number in the us
>>56807207
No thanks friend, I rather like my penis.
PS: My ancestry is celtic.
So does your data say I should be racist or not?
>>56799483
>R^2
>Negative
What the hell did you do?
>>56807322
I don't doubt that, but what you should look is percentage of whites which are poor vs percentage of black which are poor.
For example, the US Census Bureau tells me there are about 240 million whites and 40 million blacks in the US. Suppose there 10 million white poor and 5 million black poor. It is true that there are more white poor than black poor. However, 1/24 whites are poor but 1/8 blacks are poor. So the poverty rate is higher in blacks.
>>56806589
I did it a couple of years ago.
Conclusion. State level data tells us nothing.
>>56807504
My data says you should learn statistics to not be deceived by everyone and their mothers.
>>56799483
I appreciate the time you took to do this, but as >>56806141
points out, your statistics are quite flawed due to the dispersion of blacks throughout a state.
Blacks tend to congregate in urban environments. Even in states where blacks represent a small proportion of the total population, they could easily (and do in most cases) represent a large proportion of the crime in said state, which is the case in Illinois due to Chicago, California due to Los angeles...
The only logical way to approach the debate is to look at it on an individual basis, not state by state. Even a city vs city comparison would be biased because cities are going to vary based on more factors than simply racial makeup.
the fact that DC is such an outlier should be a clear sign that there is something wrong with comparing an entire state, both rural and urban areas, in one study is going to produce flawed data; DC after all is a completely urban location, which explains the high crime rate (and the high black population)
Urban areas have much higher crime rates, and much higher black populations, both of these are given. To prove that there is a correlation, you need to compare crime rates of urban vs rural blacks, and crime rates of urban vs rural whites, asians etc. I'm willing to bet that both urban AND rural blacks have significantly higher crime rates than any other race.
>>56807221
>Saying stuff like "I bet you could show a higher correlation" without actually doing the work is stereotypical racism.
No, dumbfuck, it is not. I've done the work for school performance, and it gets a lot more complex than just percentage minority or median income. There are a lot of factors that contribute to how well the students in a school will perform, and I bet that the same is true for crime.
>>56802100
This regression could use a lot of additional factors:
-Education level.
-Race (dummy variable)
-Broken home/single mother.
-Wealth (or income).
-Religion.
- Welfare recipient (dummy)
There are probably other variables which could be added that would increase the "fit" of the line.
Additionally, squared or logarithmic variables could be added to test a better "fit".
I don't buy that race is 100% behind the higher propensity of crime amongst those demographics. It may well be a factor though.
There's nothing wrong with including race, in fact, it should be considered. Ignoring potential factors does nobody any good in the long run.
Indeed, I ran a regression analysis on a data set from Africa (to try to understand the economic growth) and we basically couldn't conclude anything. Race, however, was included in the analysis.
>>56807599
https://en.wikipedia.org/wiki/Coefficient_of_determination
Summary: R^2=1-SSreg/SStot. Without explaining what SSreg and SStot are, if the model is particularly bad at fitting the data, R^2 will be negative. Basically, its like the data is behaving in a way opposite to what the model is capable of explaining.
>>56803993
I don't think you understand his point about how mostly-white rural areas with lower crime rates cancel out the extreme data found in urban areas of the same state that have high black populations.
Per capita crime rates between black urban areas and white urban areas will show a similar difference that your method of only looking at states as a single data point makes ridiculous.
>>56807741
>wants to change people's minds
>doesn't even answer a simple yes or no question
>>56807682
The appropriate conclusion in your example is that none of the dependent variables explain the independent variable. This can sometimes happen if the model is overfitted, if you have too many variables. For example, labor force participation and unemployment rate measure nearly the same thing, you should only use one of them.
>>56807750
I've said multiple time already, cities have higher crime rates because of the higher population density. Plus if city blacks behave differently than rural blacks (or whites for that matter) the different lies in the population density, not in the race.
>>56799483
Sorry OP, but you don't account for distribution or population levels.
Also the method you used to display the statistics betrays the distribution. I suggest doing a map distribution showing population / crime across the states.
>pic related; I know the second graph isn't good, but it shows Highs / Lows
I bet you think "IQ" tests aren't a good indicator of intelligence....
>>56802100
Based Russian Bro with the dubs is right on the money!
Race is an indirect factor (black males tend to beat women).
>>56807817
Good, Than do the work instead of proclaiming an a priori result. That or shut up and dont make unbased claims.
>>56803993
>My sample has a total population of 308 745 538 individuals, of which 40 250 635 self-identify as black. The total of violent crimes is 1 197 987 and the total property crimes are 8 277 829. So my sample as large as it can get without going outside the US. What you'd like me to do is cherry-pick my data. No thank you.
How is splitting up into cities cherry-picking, but spitting up into states not?
>>56807989
How the fuck could be R^2 negative. R^2 is [0,1]
>>56807843
Often you want to avoid too many variables. Overfitting tends to produce inconclusive results.
>>56808607
You should learn to actually read what the fuck people write first.
>>56808092
Then make an analysis accounting for population density and localized racial composition to prove your point instead of talking out of your ass.
Its fairly obvious that if more people live in an area, more crimes are committed. This does not imply a change in crime per capita.
>>56808200
Well, I want to change peoples minds about the proper use of statistics.
>>56803957
Whites go peacefully when they get caught. Blacks chimp out, so they end up with a list of crimes.
>>56807843
>It may well be a factor though.
Thats the entire point when /pol/ says "niggers nigging all over the place".
You're also ignoring the fact that all these variables can actually, be in fact, "interacting" with each other. If you're black, youre more likely to be on welfare, not have a HS diploma, poor, etc. just "because" you're black.
ITs like saying Asians arent short because theyre Asian, its because of low test or some shit. Well, the low test is "because" theyre asian.
Same thing with blacks but with whatever indicator they over represent as.
>>56808599
Read my five initial posts. Read my replies to the canadian dude (sorry getting tired). Also, you weren't quoting me.
>>56808609
It is cherry-picking if you do not account for all cities. What the canadian dude wanted me to do was chose a specific set of cities. For the analysis to have the validity as the state wide analysis, I'd need to account for every single city. Rural areas as well, people who live outside cities are still people and still belong in statistics.
>>56808676
See >>56807989
R^2 isn't the direct result of squaring something.
>>56809021
Except your being fucking schooled
>>56809193
You can do cross-correlation analysis to evaluate the interaction between factor, this is not an obstacle, just a lot of work.
>>56808905
I agree.
>>56809044
Is resisting arrest a violent crime in the US?
>>56807221
>if I did this my data range would lack information of regions with low black-population, and would thus be skewed.
Congratulations, you just described 20% of the data points in your own shit analysis:
you are using 50 data points, 10 of which have next to 0% black population
So no, YOU are wrong, grouping by city would NOT unrealistically skew the data as it is skewed in reality to begin win.
Did you not look at the picture I posted? Racial segregation is rampant in american society.
And when faced with this reality (as in the case of Washington DC), you merely chose to ignore it as an outlier?
>>56809325
>>What the canadian dude wanted me to do was chose a specific set of cities.
I never said just chose any specific subset of cities, that would be cherrypicking. I just gave an example of using the 10 largest cities in each state as a way to alter your analysis in order to more accurately reflect the skew in racial population %, if you want to chose all the cities you can get statistics on, then that is even better.
>>56809520
Sure.
>>56809719
Very well. You do the following. gather the data for the 10 largest cities in each state, do the analysis, post the results. Once you have those results, we can talk further, otherwise shut up and stop talking out of your ass.
can you stop being smart and shit and make chiken portugalbro? Its dinner time
>>56810164
would you like some maple syrup with that?
>>56799483
/sci/ would lynch you
>>56810356
See >>56804284
The first pic is an example of bullshit.
Well then...
Differs from year to year.
>>56799483
>No Chi Square Test
>No p values returned
>No ANOVA
lel nice HS stats
>>56811066
BTW statistics predicts correlation, not causality.
>>56811184
You have to keep things at a level that most people can follow.
I'm off for dinner.
>>56810055
>posts on pol trying to show race is not related to crime rates
>flawed analysis is pointed out
>gets given suggestions on how to make the data more accurately reflect reality
>refuses and tells people to stop talking out of their ass
Look at your data points, besides the 1 "outlier" of DC. Your x variable never exceeds 40%. Do you honestly believe that accurately reflects the skew in racial demographics that this >>56806141 post displays?
How can you not see that you are intentionally understating the dispersion in your regressor? There are parts of America that are 90-100% black, your data says no parts of america exceed 40% black.
>>56810055
You have the burden of proof you dingus
Him making the critique does not shift that in any way
Do you even peer review?
Can someone tldr me? Are we going to have mandatory penis inspections to curb black crime or not?
>>56812149
Raul is analyzing at a state level and assumes that there is no segregation below that level.
As in, he actually believes the chiraqis are spread out across the entire state of illinois and not congregated around the city centers.
He extends this belief to all US states and when reality smacks him in the face (the district of columbia) he says "MUH outlier"
>>56811241
Causality doesn't exist in soft sciences anyway.
>>56807221
> "I bet you could show a higher correlation" without actually doing the work is stereotypical racism..
wut ?
>>56808709
Of course, less variables is better. We are not talking about a simple problem though. This is not "Are unwed mothers more likely to have a lower income than wed mothers?" (for example).
The very nature of this relationship makes it complex and multivariate. Hence, extremely difficult to pin down the factors with exact precision.
>>56812149
It looks we have to target nigger crime hotspots even more than we previously thought.
>>56809193
Regression analysis attempts to tease out the various influences of each factor relative to the dependent variable.
Of course this is sketchy analysis though. Statistics, by its very nature, is a bit like a loaded machine gun- in the hands of a skilled, intelligent (and honest) statistician, it can be used responsibly and draw some good conclusions. In the hands of a chimp it can be recklessly applied.
Note: Also with statistics, the data set needs to be good. Even the best statistician/econometrist in the world will produce crap if the input data set is initially flawed.
>>56813611
Pretty much. And this is the first thing that any decent Economics/Econometrics course will teach its students at undergraduate.
Can't speak for sociology, or psychology though.
>>56806337
>but if you want to find really the root of the black problem and not just spew stormfag memes
You actually believe extremist ideologues are concerned with objective truth and don't seek simply to demonize the Other as part of an overall worldview.
>>56811895
Why are concepts like "percentage", "per capita" and basic set composition theory beyond you? None of what what cite matters to his basic analysis. In any given state, blacks have to live somewhere. As long as the crimes are reported on a statewide basis, distribution is utterly irrelevant to the conclusion being drawn from the data which is an incrediblely narrow one.
>>56800100
All that for one graph?
>>56811895
I didn't post in an attempt to show race and crime are unrelated. Rather I conclude there is a correlation >>56800644, but it appears you find it to be too low for your liking.
>>56813611
I told the guy that correlation and causality is different because his results said basketball players/fans are related to crime..
>>56811970
I need not prove anything. His desire to know the specific correlation in urban has no bearing on my conclusions on state-wide statistics. Like, I'm a chemical engineer, if I write a paper about catalysis, I have no obligation to provide evidence about the distribution of wealth in Asia.
>>56813074
I think you should read about what an outlier is and why it is excluded.
See this https://en.wikipedia.org/wiki/Outlier
>>56814323
I misunderstood what the guy was saying apparently.
>>56817351
>>56817574
>>56817771
>>56817893
btw i'm OP, just in a different IP now
>>56817351
>but it appears you find it to be too low for your liking.
Wrong, where i have an issue is how you are claiming to "educate" pol on the misuse of statistics, while blatantly misusing it yourself and when I pointed out the flaws in your analysis you responded with "thats cherry picking".
reality is slapping you in the face and you just brush it off "muh outliers"
see >>56803523
particularly the last line: "I would love to be proved wrong". You can very easily prove me wrong as well, replicate your research at a city level and come back to post your R^2 M and CI.
But I know you wont, because that will actually take work, when youd rather paste two wikipedia columns into excel, hit the solve button, make a chart and then explain why reality is an "outlier"
>>56817771
>claim: /pol/ uses statistics wrong
>evidence: a single dataset no one uses and a conclusion that is nebulous
>rebuttal: your dataset shows its flaw, by the datapoint that is not appropriate to it demonstrating its ridiculously wide net, and when that flaw is addressed by either correctly removing it or using county or city based organization the correlation is no longer nebulous but extremely clear
You either failed to demonstrate /pol/'s incompetence, or showed your own, OR more likely both
>>56818404
https://en.wikipedia.org/wiki/Outlier
you really should to educate yourself
>>56818449
https://en.wikipedia.org/wiki/Statistics
you have some catching up to do. never too late to start.
>>56818449
>>rebuttal: your dataset shows its flaw, by the datapoint that is not appropriate to it demonstrating its ridiculously wide net, and when that flaw is addressed by either correctly removing it
so we agree
in statistics, the larger the net, the better. You don't want to conclude things based on lack of data.
>>56818404
also, my sources are directly the US Census Bureau and the FBI website. You stating that I took my data from wiki shows you have not read the opening posts.
I dont see how this image is wrong
>>56818674
>constructs data set that horribly understates a regressors variance (by a factor ~2)
>reality slaps him in the face
>doesnt adjust analysis to include accurately variable distribution
>"muh just an outlier"
>"You really should educate yourself on outliers"
kek
>>56819403
Its not wrong, but its lacking the error ranges on the slope so that a confidence level can be calculated. Also there is an outlier in the four charts, the datapoint that is way up above the rest of the data cloud. Basically it alters the correlations while not really being a part of the group.
See https://en.wikipedia.org/wiki/Outlier
Assuming you have a high confidence level, you can read in bold at the bottom of the caption that the coefficient of determination R^2 the square of the correlation written in the charts, so its lower than the number in the charts.
One problem is that this is data from 2002. Demographics and economics change fast. Specifically, this is data from before the subprime bubble in 2008/2009, that event probably altered the crime landscape quite a bit.
One last problem is that cross-correlation is not provided. By that I mean how are unemployment, poverty, education and race interrelated, though this cross-correlation should not change the fundamental conclusions much.
>>56819633
you really should educate yourself
https://en.wikipedia.org/wiki/Outlier
>>56819633
let me put this way. D.C. is not a state. Period. Therefore it has no place in a state-wise analysis. However, including DC in a city-wise analysis would be perfectly fine.
Also, see >>56816853
>>56820291
you really should educate yourself
https://en.wikipedia.org/wiki/Variance
R^2 can't be negative
R^2 alone doesn't mean shit
Yes, we all know you can't relax around blacks.
>>56820450
> However, including DC in a city-wise analysis would be perfectly fine.
Yes, that is until you exclude DC, Baltimore, St Louis, South Central LA, Chicago and certain burrows of NY because they are also "outliers" in your analysis right?
you are fucking clueless raul.
>>56809685
Yes. If you resist arrest it's very easy to rack up years of additional felonies.
What us the point of statistical analysis to determine correlations between race and crime when we already have precise data on the race of perpetrators? These methods are only helpful when you trying to infer things that can't be measured directly.
ITT: Some Portunigger tries to make laughable excuses for USA niggers.
>>56820772
You are wrong, R^2 can be negative. R^2 is not actually calculated using the square of r. See >>56807989
>>56799483
>serious work
>applying simple ols linear regression, seen in any 1st chapter of econometrics 101
>>56821398
No, there no specific cities in my analysis because I used state wise data, and DC is not a state.
DC is not a state.
DC is not a state.
http://dc.about.com/od/government/qt/DCStatehood.htm
Not only is it not a state, DC is the size of a metropolis, not a state.
>>56821866
What's the range in values of R^2?
I need to know your level.
>>56821686
my ancestry is celtic
>>56809685
Sometimes. In general, it's split up int a felony version and a misdemeanor version.
>>56822014
The DOJ usually adds DC as a datapoint for these analyses, I guess because they want to include the entire country.
Would it be more correct to add DC's population, and crime to Virginia then?
>>56821866
and he didnt even do it correctly, he took the simplest population measure he could find, ignores its bias and then when the bias is shoved in his face he screams "muh outliers"
literal shitskin mathmatics 101
literally shitskin mathematics 101
>>56822416
for the purposes of my analysis, adding the population of DC to virginia or maryland would suffice. Then again I sometimes don't understand the way the census bureau works. Like Puerto Rico is a part of the US, but Puerto Rico is not in their census data. However, Puerto Rico is in the FBIs crime data.
>>56822557
https://en.wikipedia.org/wiki/Outlier
Also, DC is not a state.
I think I will stop replying to you. Unless my analysis shows what you me to show, you will proclaim I am wrong. Friend, you are either ignorant, uneducated, brainwashed or just not very intelligent and no amount of dialogue will change this.
I've found the following correlations with the violent crime rate (using state data, excluding DC):
IQ -0.316922707
Percent Black 0.516278223
Percent Hispanic 0.244909175
Poverty Rate 0.154252453
Supplemental Poverty Measure 0.539151926
GINI Coefficient 0.266432522
Unemployment Rate 0.520555835
Violent Crime Rate 1
Dem-Rep Gap -0.032992652
Obama-Romney -0.004736885
Population Density 0.164813281
Firearms per 1000 Residents -0.089409778
Firearm Ownership Rate -0.172636435
Murder Rate 0.743075103
Gun Murder Rate 0.734676452
And, using county level data (from the Census):
Total Population 0.037873093
Population Per Square Mile (2010) 0.136502303
Non Hispanic White Alone (2000) -0.447670562
Non Hispanic White Alone (2010) -0.470464883
Non Hispanic White Alone (Change) -0.150791977
Black Alone (2010) 0.419713047
Hispanic or Latino Origin (2010) 0.17267186
Violent Crimes (2008) 0.057215805
Violent Crime Rate 1
Property Crimes (2008) 0.047053435
Property Crime Rate 0.685698044
High School or Higher (2005-2009) -0.129129159
Bachelors Degree or Higher (2005-2009) 0.091773526
Persons of All Ages Living in Poverty (2009) 0.241458828
Unemployed Civilian Labor Force (2005-2009) 0.275921289
Median Household Income (2005-2009) -0.05264066
>trying to manipulate statistics ti show what you want
It isnt as complicated as you desperately want it to be
>>56822798
In the US, even if you account for all social and economic metrics, there is some amount of crime which can only be explained on a racial basis, smaller than some people like to portrait it, but its still there.
... come to think of it, testosterone levels correlated with aggression. Anyone knows if blacks have higher testosterone than non-blacks?
>>56822719
THATS WHY YOU CAN'T USE STATES YOU DIP SHIT. YOUR OUTLIER INDICATES YOUR USING A TERRIBLE DATASET. YOUR RESOLUTION IS TO LOW TO MATTER. GOD YOU ARE DENSE
>>56823360
I'm sorry but you are wrong.
See >>56816853
I think I've said this several times, but I'll cite myself again: My sample has a total population of 308 745 538 individuals, of which 40 250 635 self-identify as black. The total of violent crimes is 1 197 987 and the total property crimes are 8 277 829. So my sample as large as it can get without going outside the US. What you'd like me to do is cherry-pick my data. No thank you.
You do not agree with grouping the sample according to state, which is the largest administrative division below the level of country, and therefore the least susceptible to error. I think you are advocating gerrymandering for the purposes of obtaining whatever results you have in mind a priori.
>>56823360
>"Alternatively, an outlier could be the result of a flaw in the assumed theory, calling for further investigation by the researcher."
>https://en.wikipedia.org/wiki/Outlier#Causes
>a flaw in the assumed theory
>flaw in the assumed theory
>flaw in theory
nah yo cant be that its most likely that the shitskins in DC behave completely different to the rest of america.
Yeah, thats gotta be it right? cant be some shitskin first year applying stats 101 while criticizing others for their poor use of stats?
Nah yo, cant be that
>>56822569
It's political organization. Puerto Rico is in a weird spot. They're not in normal census data because that's used for determining distribution of seats in the house of representatives.
PR is covered by the FBI, so they need crime statistics to justify deploying officers there.
A lot of the weirdness from DC because they didn't want the local politicians getting too much control over the federal government just because they lived nearby.
>>56823259
>smaller than some people like to portrait it, but its still there.
Considering that the politically correct narrative is to say that race has zero effect on either crime rate, the inverse seems to be true as well.
>>56822798
I'm just a dumb meathead, but I have a question:
Being poor correlates with violent crime. Which is a larger influence and by how much?
>>56823794
Would your sample of exactly the same size matter if you clumped them into ONE data point? FUCKING NO. Your category of 50ish is to low a resolution and you are an imbecile
Fuck off
how much nuffin can a dindu do if a dindu dindu nuffin
>>56823259
This is the copout answer.
Crime, poverty, and unemployment are related in multiple ways. The risk of crime is an additional cost that most businesses can't bear, forcing them to shut down or move. This reduces economic opportunities, causing poverty and unemployment. The lack of legitimate economic opportunity drives many to crime.
In addition, black economic and social networks have been consistently poorer, which means they have less economic support from family and friends. The contrast is Jewish communities, who will support fellow Jews (frequently through synagogues or social networks).
While these are factors, it's more likely that blacks are simply more prone to crime, and less capable at work.
>>56824211
hes a shitskin hiding behind a thin, poorly constructed veil of "mathematics" to try and disprove the strong correlation between race and crime that is cited here on POL.
he does some really shit, flawed analysis to back up this flawed claims and when it gets shredded all he can do is say "muh outliers"
this is literal.shitskin.mathematics.101
>>56802145
NO, you simply need to do per capita and compare the numbers, and it proves that blacks do commit more crimes you do not need a bachelor is statistical analysis to understand that
>>56824472
I concur
>>56824128
Poverty alone isn't strongly correlated with violent crime (~0.15). The supplemental poverty measure, which takes into account differences in costs of living (and hence is more indicative of urban poverty than rural poverty) is much more strongly correlated with violent crime (~0.53).
However, the supplemental poverty measure is also more likely to overstate black poverty (distinctly urban) while understating white poverty (distinctly rural). Black (urban) vs white (rural) poverty is indicative of differences in population density. Liberals love to argue that population density accounts for the difference between Appalachia and Detroit. However, as you can see, that doesn't account for much (~0.16).
>>56820834
But we can relax around beautiful black women.
>>56825430
Black women aren't committing these crimes. Black men are.
>>56803523
He's not going to. It could prove him wrong.
>>56825519
I know. That's why we can relax around them.
>>56809325
He told you to choose top ten most populated per state. Nothing at all wrong with that.
hey faggot op, can i cum on your face? my data shows the more i cum on your face, the more likely you'll shut the fuck up
3rd year pure mathematics student here (no, I have not taken statistics formally but I have done data analysis using R to learn the language), the sample data chosen for this 'paper' is faulty, amongst a few other things which I won't bother to explain as I am posting from my tablet.
I tutor kids on the side, and the knowledge displayed by the OP is literally what my 16 year old student is learning at school because she took the plebeian path and went with applied mathematics.
Unemployment is the biggest contributing factor in violent crime. Poverty is actually negatively correlated with violent crime, when everything else is taken into account. Percent black and hispanic are still pretty important. This comes from census data.
>>56826608
>OP btfo by his superior countryman
>>56826562
Cum on mine instead
>>56826608
>no, I have not taken statistics
>>56823794
do you think that when people in the hard sciences try to publish faulty research and get called out on it, the acceptable response is "you try it, i dont have to prove anything."
dumbfuck here
So is OP saying that if blacks commit 20 murders per 100,000, and whites commit 2.5 murders per 100,000, they're still committing the same amount of murders when accounting for dispersion or some shit? The fuck?
A murder is a murder. And blacks are clearly 8 times as likely to commit one (in my example).
>>56825522
I don't gathering city level data is worth the work for know.
>>56826608
Out of curiosity, why don't you tell us your grade average? Just curious.
>>56828383
From >>56800644
My final conclusion is that, without looking at any other factors, the percentage of black population explain 14.2% of the violent crime rate and 12.9% of the property crime rate.
>>56828383
He's trying to compare the violent crime rate to the black proportion of the population to find a relationship. As I've shown >>56827445
unemployment is the most powerful explanatory variable for violent crime rate, while the black proportion is still very valuable. After taking everything else into account, the poverty rate is less than worthless.
>>56828379
I am a PhD student, just to clarify. I'm familiar with the review and publish procedure. Normally, stupidity is not involved.
>>56825666
You actually do need a representative sample of urban and rural populations, other you'd be looking at crimes in cities and not crimes.
>>56824735
yes you're right, but you'd need statistics to know if they commit crime because they are black or because of some other reason
>>56826608
Also, on the subject of papers, since your an undergaduate, have you ever written or publish a paper?
>>56828830
Crime in cities would still be valuable, since rural areas experience less crime simply due to the lower availability of victims. Population density is one explanatory variable of crime, which is controlled for (partially) by only including cities.
>>56829164
Or you could see population density as a variable and use data from a range of densities to access it.
>>56825633
Exactly. And because this is thanksgiving, I should say that I am thankful for God creating these beautiful creatures.
Amen.
>>56826608
Also, see >>56811241
You know, at some point you realize that the average man won't really be able to understand you unless you simplify your explanation.
>>56829426
What the fuck is that. Look at that nasty butt.
>>56799483
Thank you for the statistical work, OP.
Nigs gonna nig though unfortunatly.
>>56829598
>What the fuck is that
God's creation you fucking apostate.
>>56829692
Surprisingly, people have been complaining that the correlation between crime and race I'm presenting is too low for liking.
Its like feminazis, if you acknowledge there is some truth to their complains, they'll be even more mad.
Bed time. OP out.
PS: I think some of the people in this thread are just butthurt cause they got mugged by some black dude recently. Get over it, shit happens.
https://www.mensa.se/provtestet/provtestet.html
Ya faggots better score >126 or be embarrassed.
>All questions are figures/matrices.
>Time limit = 10 minutes.
>Test standardized for 18 and up.
>Starta Testet
>>56829883
thank you for this thread and your work based mathemagician
>>56829426
I would go bananas over a body like that.
Unfortunaley, I live in Wales.
>>56830302
Is this like every other feel good test on the internet or does mensa give it some credibility?
>>56831826
>Unfortunaley, I live in Wales.
My condolences.
>no guns
>no black women
Real shame.
You're very ambiguous as to what data you're using, is it violent crime, property crime or what?
Also using state-wide data is an extremely poor decision as many states contain areas with both high concentrations of blacks and those without, it will inevitably muddle the result.
>>56832419
https://en.wikipedia.org/wiki/Mensa_International
Scoring well on the (test)test basically equates to:
>You're not retarded - please contact us in order to conduct a supervised test.
If someone doesn't max out their online test, they shouldn't hold their intelligence in high regards.
>>56832419
>>56835396
And they charge you money for it, so I wouldn't doubt that it's a bit generous with the scores. The more suckers they can lure in the better.
>>56830302
>pattern recognition
>reliable method of measuring intelligence
>>56835396
>>56835561
I mean the online test, the actual test taken in person is considered fairly accurate but what about their online one. I guess I should retake it and give random answers. If the results say I'm a genius then we can assume the test is bullshit.
>>56828301
Not having formally taken statistics doesn't exempt me from knowing it. It's applied mathematics, it's not hard.
>>56828503
>Out of curiosity, why don't you tell us your grade average? Just curious.
I go to FCUP, and no, I won't specify it. Just that it's >16.
>>56829112
Published? No, never bothered to try. Written? Yes, I have written a few proofs that were peer reviewed. Namely, in geometric topology, surgery theory, and galois theory.
Off to bed folks. Have a good night.
>>56835561
>>56835870
Just did a few test with random answers and it gave me a result below 75, which is as low as it goes. So the test appears to not be complete bullshit.
Anyone with half a brain can get max score on that test, on a side note i don't understand all the fuzz, from what i can gather from op calculations you can see that niggers commit at least 2x the crimes whites do
>comes to pol trying to school us on stats
>post flawed methodology
but what about DC?
>"HURR NO THATS DIFFERENT: the brown people in DC act COMPLETELY different to those everywhere else in america because its not a state, its a district, HURR OUTLIER"
Maybe its not an outlier raul? maybe shitskins dont behave differently because they living in a district and not a state
>"HURRR OUTLIER"
>>56801809
>In statistics, the larger the sample, the more reliable the results. Scaling-down from the state level to the city level decreases the sample size.
That's wrong. Well, the statement is sensible but using cities instead of states would give you more observations, not less, and would better display the result since cities or preferably districts are a lot more segregated than states. So in this case the statement is just plain wrong. Cities would give more observation(reliability) as well as more exact individual measurements.
>>56801809
>Scaling-down from the state level to the city level decreases the sample size.
Oh look, it's a retard trying to apply his statistics class and failing.
>>56799483
lol you retard r square can neither be negative or bigger than 1. you fail miserably
>>56838045
>Oh look, it's a retard trying to apply his statistics class and failing.
Yeah, I have no idea why he even included states without without a sizable black population. Blacks only misbehave when they segregate and have to live with other blacks.
In Portugal, the blacks we have up north are quite nice, as there are very few of them, but the ones in Lisbon and AMADORA are the devil (especially the latter).
So yes, there is a condition that must be met for blacks to resort to crime: self-segregation and the creation of ghettos.
Yes, I am aware whites commit significantly less crimes in mostly white ghettos, no need to point that out, it's just how we are.
>>56838178
>lol you retard r square can neither be negative or bigger than 1. you fail miserably
Wrong. Fitting non-linear functions to data may result in a negative coefficient of determination.
>>56838994
>So yes, there is a condition that must be met for blacks to resort to crime: self-segregation and the creation of ghettos.
OP, I propose you repeat the study, but divide it in the following categories:
[0-5%]
[5-10%]
[10-15%]
[15-...]
Based on https://en.wikipedia.org/wiki/List_of_U.S._states_by_African-American_population
>>56808499
>cities have higher crime rates because of the higher population density. Plus if city blacks behave differently than rural blacks the different lies in the population density, not in the race.
That's an assumption. Which, if false, could explain the discrepancy in your analysis and everyone else's in part or even completely.
And stating an assumption, even "multiple times", does not make it true.
>>56833960
I'm not really fussed about guns but I do feel for my euro bros concerning the muslims at the minute. They could do with some guns.
Curvy women of all ethnicities are most welcome though! All I've got to choose from are no curves women and sheep.
>>56839783
>That's an assumption. Which, if false, could explain the discrepancy in your analysis and everyone else's in part or even completely.
True. He's wrong, because he didn't remove that factor from the equation, so to speak. If the problem lied in the population density, then why do blacks commit crime at substantially higher rates, when accounting for population density (eg: compared to whites living in the same city, under the same circumstances)?
Alright, this is my last post, I am going to sleep for real now.
what the fuck does it mean? this is too complicated. do blacks commit more crime or not?
>>56820601
why are you arguing with an inbred fish eating peasant?
>>56807221
> I should look at the specific locations with the highest concentration of black population
You sure shouldn't exclude them.
>The appropriate method is to look at crime per capita across the largest range possible of black density, so that I can access the importance of black density in crime
The "appropriate method" is not to expand every individual observation. You want a large total population in the entire study, not in every individual data point. That muddles the result.
>thought experiment
Lets say we have an entirely white state, save for a large black city. Outside the city it's safe, there is hardly any crime. While the black city looks like fucking Iraq. The state would then have some medium-low amount of crime with a pretty low but sizeable black population. In this scenario too wide an observation(whole state) would miss the problem entirely.
You've used a method, which assuming the result you're examining is true(blacks=crime) would still fail to show it. That's no good.
>>56839941
Well that could be true as long as black mainly lived in cities and cities have a lot higher crime per capita. Then white figures would look low in comparison, dragged down by the nice rural folk who wouldn't be significant in the black population.
honestly tldr.
I know i could understand all the calculations you did but all you have to do is live around american nigs for a good period of time to learn that they are fucked up in the head, and commit a disproportionate amount of crime
>>56800100
That shit approaches unity.
>>56840877
He uses entire states as individual data points and does a linear regression. And given how large the states are he finds fuckall correlation.
>>56841033
using states isn't a very good metric. Considering there are many states that have a lot of white people, where a great deal of the crime are committed in inner city black neighborhoods.
hell my home state is almost all white, but everyone knows to never go to a certain part of town because of all the crime. That area also happens to be the part where all the blacks live.
>>56808282
>The appropriate conclusion in your example is that none of the dependent variables explain the independent variable.
the dependent variables can't explain the independent variable by definition, retard
>>56841381
Exactly, and that's why his study has such a remarkably low correlation between the proportion of blacks and rate of crime. Also he threw out D.C as an outlier.
>>56799483
One way of interpreting your statistics is that blacks will:
* commit more crimes were there is more to steal/rob and that there is more to steal the more non-blacks that live in an area.
* that the higher the black part of the population is, the more likely it is that the perpetuator gets shot by the victim, and that this is a deterrent.
Your analysis does not demonstrate that blacks aren't responsible for more of the crime
>>56840175
>You've used a method, which assuming the result you're examining is true(blacks=crime) would still fail to show it. That's no good.
/thread
OP was a faggot yet again
>>56821610
I'm still waiting for somebody to answer my question
>>56813074
I just want to help out your bigotry with some of my own. A Portuguese dude that can speak english? His name is Joe. Their name is always Joe. Maria for a chick, but that's like 80% of the women anyway. Also I think I just triggered OP with my half-assed use of a guess of a statistic there.
Someone do linear regression on Islam and violence, do one for middle east and one for north America
>>56799483
If your subjects fall into distinct groups, and if these
groups tend to have different average values of X and Y,
then correlations may arise simply because of group
differences.
13 states are over 15% black. 19 states are under 20% black. 4 states are over 30%. 4 are under 1%.
Yet, you group all of these states. Do it correctly.
See: >>56839163 for a decent idea of how to do it properly. I'd suggest different categories for < 1% and greater than/equal to 30%.
>>56803236
Why are you using a simple linear regression model when there are other obvious and easily obtainable statistics that relate to crime rates.
![t8EI6cL[1].png](https://i.imgur.com/BzTUZLA.png "t8EI6cL[1].png")
>>56801809
>This shows qualitatively that there is cross-correlation between crime, race and income.
No, the chart doesn't include a correlation measure at all.
It's a visual analysis of what realm of poverty and crime have a given distribution for a certain demographic. It's only slightly better than using a visual overlay of data on a geographic map.
The interpretation for you is hinging on your bias as to what the resulting correlation or causation might be for the displayed data.
Being poor increases likelihood to commit crime. But the rate at which this change in income affect the crime rate varies widely for each demographic.
Hispanics for instance have the same median income as Blacks, but the Hispanic crime rate curve over income trends is MUCH flatter.
The curves also change slightly for each demographic depending on population density.
http://www.bjs.gov/content/pub/pdf/hpnvv0812.pdf
These factors of income and population density are manifesting themselves to different degrees among the different demographics. It seems the absolute worst-case scenario is the be being Black, living in an Urban area, and Poor. In that order specifically.
This particular chart was originally made as a political spectrum visualization.
