3 ms·

My point was simply that there is not agreement across the board that they commit more crimes. Yes, people are unscrupulous when it comes to statistics, that's

My point was simply that there is not agreement across the board that they commit more crimes. Yes, people are unscrupulous when it comes to statistics, that's definitely true. But "adjusting" for things is usually a way for people to get the result they want with the numbers. The government does this all the time with "seasonal adjustments" to housing and employment data. There never appears to be any rhyme or reason to the adjustments other than getting a better/desired number. Also, when people argue that it's because of socioeconomic factors or whatever else they may claim, that doesn't change the fact that they either do or do not commit more crimes. The reason for it can be discussed and addressed but many people try to use such an explanation as a way to weaken the fact that is presented. I think the general citizen needs to become far more knowledgeable about math and statistics since they are always used to justify public policy nowadays. The common person is terrible at being able to weed out the bullshit that is presented by sources which all have agendas.

dragonwriter 10y ago
> But "adjusting" for things is usually a way for people to get the result they want with the numbers. The government does this all the time with "seasonal adjustments" to housing and employment data. There never appears to be any rhyme or reason to the adjustments other than getting a better/desired number. While "adjusting" may be used for misleading purposes, government seasonal adjustments generally used a disclosed-in-advamce methodology with a clear purpose; they only seem to have no rhyme or reason if you don't bother to do even the most basic checking. An overview of seasonal adjustment at the Bureau of Labor Statistics is available at [0], and more detailed information about the specific methodology and factors used in particular series are readily available. [0] https://www.bls.gov/cpi/cpisahoma.htm https://www.bls.gov/cpi/cpisahoma.htm
- turc1656 10y ago
  I disagree. That link shows no formulas whatsoever, just a generalized explanation of what they do. Without a hard set of rules to follow, it's useless. Also, looking into the X-13 ARIMA software documentation, it looks similar to other statistical software packages like Eviews, which I have used in the past. This methodology they follow still highly subjective based on the parameters used in the model. This is very much like what my employer does for publishing public documents for financial compliance. We are required to publish a public methodology/rulebook for our products so that investors can, in theory, replicate everything we are doing on our side. In practice, though, forget it. No one can replicate it. The document serves to explain the concepts behind the product and the general guidelines, but no one can fully know the granular details that make all the difference. It sounds like the same thing is going on here. No one can realistically be expected to take the same gigantic set of data, individually set a dozen or more parameters and run it through software and get the same result. And that's not even counting data sets that the "clean" by discarding certain outliers before running the model.
  - dragonwriter 10y ago
    > I disagree. That link shows no formulas whatsoever, just a generalized explanation of what they do. I said the link was a general overview of seasonal adjustment used in data series by the BLS, and that they provide additional detail the about the specific methodology used in specific series elsewhere. But the key part from that summary relevant to your comment upthread is that not only are the specific methodologies for particular series published in advance, but the resulting adjustment factors for future time periods are determined and published in advance, before the data to be adjusted exists. One may disagree with the methodology or the figures, but neither is the methodology opaque nor are the factors chosen to get desired results (since the actual raw data, and thus the factors necessary to get produce desired results, are not available.)
alkonaut 10y ago
If you didn't adjust, and immigrants somehow still were convicted of fewer crimes then that would be truly astonishing since then the poor/unemployed immigrants would be committing vastly fewer crimes than the corresponding group among non-immigrants. Doing these kinds of adjustments is not dishonest if done correctly and in these circumstances it's often absolutely necessary to produce useful stats. How stats are to be massaged depends on what argument one tries to make though. There are valid arguments from uncontrolled data too. An example could be e.g "the immigrants we allow in are going to be unemployed and poor which means we are going to get more crime, I don't want that so I'm against immigration" - this is an entirely valid argument to make from uncontrolled data. This is a two step argument: immigrants are poor and poor people more criminal. This is not the same as saying "immigrants commit more crimes" because if unqualified such a statement always assumes controlling for relevant factors.
- turc1656 10y ago
  I agree that controlling for certain factors is a core principal in statistics so that you can tease out relationships once you quantify the impact of other factors. Also, "if done correctly" is a huge assumption with everyone having an agenda in the current climate. But what I was trying to get at is not that the argument should just be with that one uncontrolled data point. My issue is that society can't even agree on that basic, top level data point to arrive at the first-stage fact of whether immigrants do/don't commit more violent crime. Everyone would have to agree on that first, then they can dig into the "all else being equal" scenarios where you control for the other factors that influence crime.

dragonwriter 10y ago

> But "adjusting" for things is usually a way for people to get the result they want with the numbers. The government does this all the time with "seasonal adjustments" to housing and employment data. There never appears to be any rhyme or reason to the adjustments other than getting a better/desired number. While "adjusting" may be used for misleading purposes, government seasonal adjustments generally used a disclosed-in-advamce methodology with a clear purpose; they only seem to have no rhyme or reason if you don't bother to do even the most basic checking. An overview of seasonal adjustment at the Bureau of Labor Statistics is available at [0], and more detailed information about the specific methodology and factors used in particular series are readily available. [0] https://www.bls.gov/cpi/cpisahoma.htm https://www.bls.gov/cpi/cpisahoma.htm

turc1656 10y ago
I disagree. That link shows no formulas whatsoever, just a generalized explanation of what they do. Without a hard set of rules to follow, it's useless. Also, looking into the X-13 ARIMA software documentation, it looks similar to other statistical software packages like Eviews, which I have used in the past. This methodology they follow still highly subjective based on the parameters used in the model. This is very much like what my employer does for publishing public documents for financial compliance. We are required to publish a public methodology/rulebook for our products so that investors can, in theory, replicate everything we are doing on our side. In practice, though, forget it. No one can replicate it. The document serves to explain the concepts behind the product and the general guidelines, but no one can fully know the granular details that make all the difference. It sounds like the same thing is going on here. No one can realistically be expected to take the same gigantic set of data, individually set a dozen or more parameters and run it through software and get the same result. And that's not even counting data sets that the "clean" by discarding certain outliers before running the model.
- dragonwriter 10y ago
  > I disagree. That link shows no formulas whatsoever, just a generalized explanation of what they do. I said the link was a general overview of seasonal adjustment used in data series by the BLS, and that they provide additional detail the about the specific methodology used in specific series elsewhere. But the key part from that summary relevant to your comment upthread is that not only are the specific methodologies for particular series published in advance, but the resulting adjustment factors for future time periods are determined and published in advance, before the data to be adjusted exists. One may disagree with the methodology or the figures, but neither is the methodology opaque nor are the factors chosen to get desired results (since the actual raw data, and thus the factors necessary to get produce desired results, are not available.)

alkonaut 10y ago

If you didn't adjust, and immigrants somehow still were convicted of fewer crimes then that would be truly astonishing since then the poor/unemployed immigrants would be committing vastly fewer crimes than the corresponding group among non-immigrants. Doing these kinds of adjustments is not dishonest if done correctly and in these circumstances it's often absolutely necessary to produce useful stats. How stats are to be massaged depends on what argument one tries to make though. There are valid arguments from uncontrolled data too. An example could be e.g "the immigrants we allow in are going to be unemployed and poor which means we are going to get more crime, I don't want that so I'm against immigration" - this is an entirely valid argument to make from uncontrolled data. This is a two step argument: immigrants are poor and poor people more criminal. This is not the same as saying "immigrants commit more crimes" because if unqualified such a statement always assumes controlling for relevant factors.

turc1656 10y ago
I agree that controlling for certain factors is a core principal in statistics so that you can tease out relationships once you quantify the impact of other factors. Also, "if done correctly" is a huge assumption with everyone having an agenda in the current climate. But what I was trying to get at is not that the argument should just be with that one uncontrolled data point. My issue is that society can't even agree on that basic, top level data point to arrive at the first-stage fact of whether immigrants do/don't commit more violent crime. Everyone would have to agree on that first, then they can dig into the "all else being equal" scenarios where you control for the other factors that influence crime.