10 comments

[ 3.6 ms ] story [ 38.1 ms ] thread
If only the data underlying this visualization were complete.
The data is useless. My home city has had >0 OIS during the date range, but the data reports zero.
Clicking through randomly, it seems that the armed victims were less likely to be killed than the unarmed victims. I wonder if the underlying data refutes that observation.

           Killed   Not Killed
Armed________729______300

Not Armed_____211_____114

The Chi-square statistic is 4.0812. The P value is 0.043363. This result is significant at p < 0.05

Yes, armed vs unarmed were different, question is in what direction. IOW were those armed more likely to be killed than the unarmed?

For armed "shootees", killed/not-killed == 2.43, whereas unarmed == 1.85.

Looking at it the other way, for those killed, ratio of armed/unarmed was 3.45, but only 2.63 for shooting survivors.

Since the categorical data was shown to differ significantly, the difference ratios are also significant. We can conclude one is more likely to be shot dead if armed when confronted by police.

Edit: grammar

Suggestion: try semi-transparent bubbles for each data point. It will create a heat map which is faster to read and understand than summaries per region.
See Framingham Massachusetts. Same victim listed 3 times.