Hacker News new | ask | show | jobs
by quantumduck 1490 days ago
I mean correlation doesn't mean causation. Perhaps the very reason the bought a handgun is because they thought they had some threat or were involved in activity that would increase their likely hood of getting killed.
1 comments

This study looks at the broader question of people who started living with handguns in the house, from a baseline of no handgun, so many of those victims are not people purchasing weapons themselves.

The absolute numbers are low (12 killings / 100,000 pop as opposed to 8 / 100,000 in the control group).

I don't think the article claimed causation but you can't determine causation from this type of study either. There's a possibility of the data being random, as well as causality going the other way, but it is hard to temporalize getting killed causing someone to start having a handgun in the home before that point. There's also a possible third unknown and causal association that is responsible for both the killing and the prior presence of the gun. That is a definite place to do research.

If we start with those numbers and a null hyp that both the control and the "experimental" group have a mean of 10, then the chance that of finding 4 or more extra killings than in the "experimental" group is about 21%. So even assuming no p-hacking, your p-value is meaningless.

R-code:

  sum( (rpois(10000,10) - rpois(10000,10) ) > 3)