Wednesday, March 21, 2018

More gun control ideas destroyed by crummy data

In line with the awesome analysis found by T-Bolt (highly recommended) comes this via Isegoria: How does the number of steps required to buy a gun relate to homicide rates?  Short answer: not really at all.
Not what I wanted to do this morning, but when I saw a fellow sociologist Tweet about a New York Times story on “How to Buy a Gun in 15 Countries,” I couldn’t help myself. According to the Times, “Many Americans can buy a gun in less than an hour. In some countries, the process takes months. Here are the basic steps for how most people buy a gun in 15 of them.”
The implication is that adding more steps and required approvals to the process of buying a gun ("Be more like Europe!").  The problem is that it doesn't.  Looking at the homicide rates in other countries vs. how many steps and approvals are required to buy a gun shows no correlation.

The red dots are the data points for each country and while a "best fit" line has been plotted, the bit is very poor.  The "R2" value in the upper right is exceedingly low; normally you need R2 to be at least 0.5 for a valid correlation and mostly you want to see R2 > 0.7.  This value is 0.071, meaning that there is basically no correlation at all between the data points.

Combining homicide and suicide rates, the closest countries to the USA (2 steps, 14.58 combined rate) are Austria (8 steps, 12.61 rate) and Yemen (2 steps, 16.67 rate).

Summing up: The New York Times says that we need to have more restrictions and permission steps to buy a gun ("Be more like Europe!") when this would have precisely ZERO effect on the homicide or suicide rates compared to Europe.  In other words, it's just another comfortably smug northeasterner blabbering nonsense about a topic he knows nothing about.


SiGraybeard said...

Wait ... so in this guy's world, a correlation of .071 is meaningful enough to be a "weak correlation"? When I took stats, that was considered just random variation. And the best one here, 0.123, was as well.

Wake me up when the correlation is over 0.75. I won't even begin to think of a relationship until that.

Borepatch said...

Heh. The action for a lot of stuff in in the R2 number. Google "Caspar and the Jesus Paper" for an incredible example of this in Global Warming science.

Yeah, a value of 0.071 is a big flashing neon sign saying NOTHING TO SEE HERE, MOVE ALONG

Eck! said...

I find that laughable...

First in NYC you can't buy a gun at all and if you could its likely
to take more than an hour to get to an out of NYC store...

Here in MA you must take a safety course first then get a FID or LTC-A to buy any gun and some require the LTC-A. You can't do that in a hour.

Then is 4473 form and NICS. The chart is noise and lies, the bigger lie is what they leave out.


Richard said...

It is useless to talk about data. They don't care and they are not interested in reducing homicide rates. The goal is to disarm law abiding citizens.