I saw a thing on TV about how there's a state by state correlation between gun ownership rates and gun related deaths. I don't agree with doing it that way because part of the question is whether people would just find other ways to do bad things if there are fewer guns. I think one should use something like homicide rates.
So anyway I had time on my hands today and decided to just analyze some data just like I would analyze environmental data per my job. I got 2016 homicide rates from CDC. I got state by state gun ownership rates at https://www.cbsnews.com/pictures/gun-ow ... -state/51/. I got race and poverty rates data at the Census site. The years don't exactly match but I think the data should still reasonably correspond. Don't think, for example, the State by State gun ownership rates should've changed must between 2015 and 2016.
Anyway, I started off just wanting to look at race so I looked at both % Black population and % non HIspanic White population. One is just looking at Blacks and one is basically a reverse way of looking at everybody except non Hispanic whites. Here are the correlation coefficients for state by state homicide rates vs. each variable I looked at:

The way it works is that the higher the absolute value the stronger the evidence for correlation. As you can see the positive correlation for % Black, about 0.726, has a higher absolute value than the negative correlation for % White. So I picked % Black to represent race.
First off, in order for a correlation in this set to be "significant" by convention, the absolute value would have to be [>0.279. So right off the bat there's not sufficient evidence to say there's a correlation between gun ownership rates and homicide rates. But I went ahead and did the same kind of multivariable modeling I do in my job to see what would happen if I tried looking at gun ownership rates while taking race and poverty into account. You know, see if I could tease out a gun ownership rate effect.
Nope. To make a long story short...or at least a long part of this long story short...the model I ended up with says that if you want to try to "predict" homicide rates of states you would use % Black population and poverty rate. Gun ownership rates wouldn't factor in at all.










