4 ms·
> There are no studies showing that percentage of people carrying a firearm is in any way connected to the crime rate. Once you account for all the other variab
by AnthonyMouse 8y ago
> There are no studies showing that percentage of people carrying a firearm is in any way connected to the crime rate. Once you account for all the other variables, there isn't any correlation.
Social science statistics are terrible. If you look at the statistic without controlling for anything, it's wrong. If you try to control for other factors, you can generally make the result be whatever you want.
There are multiple overlapping reasons for this.
One is that statistical controls, like statistics in general, don't reveal causation. If you want an effect to disappear, find something that correlates with the outcome and use it as a control. The effect disappears regardless of whether the control causes the outcome. The outcome can cause the control, or they can both have a shared cause, and you can still use it to erase the original effect. For example, do the demographics of the area affect the crime rate, or does the crime rate affect the demographics of the area?
There are also arbitrarily many things that could be controlled for. Taking some of them into account will make the original relationship weaker. Taking others into account will make the original relationship stronger. Looking at only the ones that move the needle in the direction you want it to move lets you move it in that direction.
If you really want to make something inconvenient disappear, just use the metric as the target. Audi claims its vehicles are safer than some competitors. Partisans collect data from two similar areas with a different rate of Audi ownership. The number of fatalities involving Audis is higher in the area with a higher number of Audis. Does this disprove Audi's safety claim?
When you have a politically charged issue like this, you can't pretend that one side's numbers are true and the side's aren't, when the truth is they're both biased and therefore both invalid.
This is a strong candidate for an adversarial collaboration, but to my knowledge this has not actually happened.
- int_19h 8y agoIn this particular case, the thing is that both political sides have tried to do studies to prove their point. Most of those studies, as you say, are too biased to pay any attention to - they either cherry-pick data, or conveniently ignore some unrelated variables that could reasonably explain the effect, or involve convoluted models with a lot of subjective assumptions in them. But every now and then, you get studies that don't do any of that. And then a funny thing happens: correlation disappears, regardless of who commissioned the study. Both sides vehemently object to these findings, which to me is a good sign that they are closer to the truth than anything else. I am pro-armed-self-defense personally, and regularly carry concealed. But it would be a delusion for me to think that by doing so, I contribute to crime reduction.
- AnthonyMouse 8y ago> But every now and then, you get studies that don't do any of that. And then a funny thing happens: correlation disappears, regardless of who commissioned the study. Both sides vehemently object to these findings, which to me is a good sign that they are closer to the truth than anything else. Neutrality bias is actually still bias. A lot of honest people do this by mistake, thinking that there should be balance and then seeking out things that create it and disregarding things that don't. But when A says +5 and B says -5 and the truth is +3, the truth is still +3, not zero. Sometimes things really do balance out exactly, but it's not very common. Round numbers are inherently suspicious.