I haven’t done the math, but I would have thought that a hundred incidents would be more than a hundred times as much evidence as one, because it says that it’s not just the unsurprising lunatic fringe of your supporters who are up for violence.
I don’t think that’s possible, unless the first incident makes it conditionally less likely that the second incident will occur unless Hitler is ungood.
Unless you mean, “the total information that a sum of one incident has occurred is less than a hundredth the evidence than the total information that a sum of a hundred incidents have occurred”, in which case I agree, because in the former case you’re also getting the information on all the people who didn’t commit violent acts.
unless the first incident makes it conditionally less likely that the second incident will occur unless Hitler is ungood.
That wasn’t what I had in mind (and what I did have in mind is pretty straightforward to express and test mathematically, so I’ll do that later today) but it’s a possibility worth taking seriously: are you the sort of organisation that responds to reports of violence with a memo saying “don’t go carving a backwards B on people”?
Assuming the prior probability of politically-motivated violent incidents to be greater than zero, X incidents where X/(number of supporters) is roughly equal to the incidence for the entire population offers very little evidence of anything, so X*100 is trivially more than a hundred times the evidence.
I haven’t done the math, but I would have thought that a hundred incidents would be more than a hundred times as much evidence as one, because it says that it’s not just the unsurprising lunatic fringe of your supporters who are up for violence.
I don’t think that’s possible, unless the first incident makes it conditionally less likely that the second incident will occur unless Hitler is ungood.
Unless you mean, “the total information that a sum of one incident has occurred is less than a hundredth the evidence than the total information that a sum of a hundred incidents have occurred”, in which case I agree, because in the former case you’re also getting the information on all the people who didn’t commit violent acts.
That wasn’t what I had in mind (and what I did have in mind is pretty straightforward to express and test mathematically, so I’ll do that later today) but it’s a possibility worth taking seriously: are you the sort of organisation that responds to reports of violence with a memo saying “don’t go carving a backwards B on people”?
Assuming the prior probability of politically-motivated violent incidents to be greater than zero, X incidents where X/(number of supporters) is roughly equal to the incidence for the entire population offers very little evidence of anything, so X*100 is trivially more than a hundred times the evidence.