Zero correlation of Brady Center scores with (preventing) violence

I read the other thread about the Brady bunch scoring states on gun control. Well, that gives a pretty good starting point for seeing whether their measures work. If the Brady center is right, then states that adopt more of their measures should see violent crime reduced. Note that I'm not graphing gun crime, because I honestly don't care whether I get killed with a knife or a gun. Done either way. I'm interested in the overall effect of Brady legislation on violence. The sources are included in the graphs, which came out a little bit big.

The result: effectively zero correlation. The R2 value shows how strongly they correlate. 1.0 is perfect, 0 is no relationship. The correlation between brady scores and crime rate is .000093, but in the wrong direction: states that have higher brady scores have very very slightly higher crime rates.

Now, what if there was some other thing that might reduce violent crime?

I also graphed % of population with concealed handgun licenses against violent crime. There turned out to be a slight correlation, just under 0.1. It's weak, but arguably present. Both graphs are below. My apologies again for the over-sized images.

http://bayimg.com/image/jammiaacc.jpg

http://bayimg.com/image/jammmaacc.jpg

re: beaker-- statistical stuff, technical

Quote:

Originally Posted by

**beaker**
wow. a blast from the past there. Hopyard bringing back things i never thought i would ever need to remember in math. i think you're right about the % thing. for it to be correct the % of concealed carriers would have to be concealed carriers per 100,000 population.

and wouldn't R^2 represent a parabolic graph. its a linear graph so yeah, it should be just 'r'

I like the arcsine transformation because it is quick and easy and some programs will do it for you. You are right I think that rate (carriers per 100,000 population) could be used in stead of %

I don't know about what a graph of r squared would look like, but I am certain that the statistic for correlation coefficient is r and not r squared. This actually makes your results look much better than the way you presented them. An r of .313 ain't bad in real terms if there are lots of 'n's present in the whole population.