re: #429 Hal_10000

See my comment above. I fit their data as a function of time. There is a small slope (about 1.2 extra shooting victims per year) but it is also statistically consistent with a flat trend.

Great. So run a one-way anova, which is a lot more appropriate test of variance than a slope—why the fuck are you using a slope?

Have the dependent variable be ‘days between shootings’. Have the independent variable be ‘post-2011’. You’ll find, as I did, that there’s a p<.05 that the data from 2012 onwards are different from 1989-2011. Yes, the N is less than a hundred so that’s criticizable right off the bat, but with the available data, this is the best that can be done.

I do thin it’s likely that what we’re seeing is not a steady ramp up that will proceed into the infinite future, but a cluster. How likely that is would require extremely complex analysis of the entire data set, not just eyeballing it like you are. It is ridiculous that you are criticizing data as weak by just casually doing back-of-your-eyelids calculations like this. Also, say that stuff does ‘cluster’. Is that because of randomness, or because of actual environmental factors? You seem to be assuming it’s random chance, but it may very well have to do with some environmental issue and it may very well not reduce until that issue has been changed.