[ekjhgkejhgk]

> “Overall, 76% of female murders and 56% of male murders were perpetrated by someone known to the victim.”

> https://bjs.ojp.gov/female-murder-victims-and-victim-offende...

Lets say M is "being murdered" and A is "stranger in the house", "not A" is "person known to the victim in the house".

The numbers you're quoting say that P(not A | M) is large, implying that P(A | M) is small.

However, to make a decision on whether to let someone in, I care about P(M | A).

You need to exercise that critical thinking more. You just heard someone say "the murders are known to the victim" and you instantly dropped your common sense.

[card_zero]

I don't think statistics are relevant at all. Suppose the stranger is wielding a kopesh, an ancient Egyptian sword. What we want to know is not "how many murderers use kopeshes?" (none of them), but "is this guy a murderer?", and that seems in line with what you're saying about statistics. However, the question "how many wielders of kopeshes are murderers?" is also irrelevant, and the answer is still none of them. Similarly, "how many strangers in your house have been murderers?" is irrelevant, even if the answer is "all of them so far". Perhaps you only ever let one stranger into your house, and once inside she killed somebody with an arquebus, and you said "never again" - but that would be paranoia. Perhaps you look at country-wide statistics for the average stranger (these aren't kept), but you are not personally country-wide, and the specific stranger is not an average. What's more, if you befriend the stranger, what statistic do you want to use then? The thing to do is reason, not count. I think the 76%, 56% statistic (although irrelevant to a decision) is attempting to say a lot of murderers are motivated by interpersonal relationships, you know, and get you to think about what a given person might be up to, or might want, and the extent to which you can even tell, and the value of risking the unknown.

[ekjhgkejhgk]

Nothing that you said prevents one from discussing models and making estimates.

All you're saying is you don't like my model (presumably because you'd like more inputs to the estimation?). Ok. You might not like my model, but at my comment on conditional probabilities was correct. The person that I was responding to wasn't that.

[scotty79]

> Similarly, "how many strangers in your house have been murderers?" is irrelevant, even if the answer is "all of them so far".

That doesn't sound very sane.

[ekjhgkejhgk]

Complete word salad.

[scotty79]

I think what he was trying to say is that the next coin toss can't be predicted based on previous one, which is true. However history of results up to now can tell a lot about where you'll end up and with what probability if you keep tossing the same coin. And that's what really matters in life. Even in real world gambling.

[bluechair]

Damn. Amazing response.