News for hackers. Stuff that matters.
Re:Facts about migration and crime in Sweden by alkonaut on Friday February 24, 2017 @01:15PM (#13726083)
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.

Re:Facts about migration and crime in Sweden by turc1656 on Friday February 24, 2017 @01:50PM (#13726447)
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.