And you're doubling down by comparing whole apples to orange sections.
If I'm going to use statistics to rationally calibrate my fear emotion, I should ensure that I am using the appropriate statistic.
I'd start off with the assumption that most violent crimes against me will be committed by another person that is both aware of my presence and able to make a credible threat of force against me. I'm probably not going to get punched in the face by anyone who cannot reach my face with their fists.
And I can further break that down by the fraction of time any given person presents me with the threat of violence. I may, perhaps, use risk of violence per minute of exposure as my metric.
It immediately becomes obvious that the publicly-available published crime statistics are useless for my purposes. They do not contain sufficient information to allow me to determine how much more or less likely I am to be attacked by a person of a given skin color for each minute of exposure to that person in public. I would have to know detailed social habits for everyone within the crime statistic reporting area.
It may be that for every 100000 opportunities to punch someone in the face, the conversion rate is higher for one skin color than for another, and I should therefore rationally calibrate my fear accordingly. But I really have no way of accurately determining that useful statistic from published crime data.
But even with that hanging over us, I still have to object to your misuse of statistics. You said:
> You are 4 times more likely to be violently attacked by
> a white person than a black person, if you are a white person.
And this does not follow from the previous paragraph. Given a randomly selected white person who was attacked in the last year, it is 4 times more likely that the attacker was white than black. Manipulating that conditional probability to find the ratio of probabilities that a randomly selected white person was attacked by a white person compared with a randomly selected white person being attacked by a black person requires additional Bayesian calculations that you did not perform. You would also need to know population sizes and overall violent crime rates.
But really, this is all moot, because fear is not rational.