Posted: Jan 08, 2015 5:16 pm
Thommo wrote:Lowpro wrote:I have a mostly similar position but one difference. If you do not have data for a position (and I mean data in a strict information stance) then you have NO POSSIBILITY. A probability can be inferred with a sample size of 1, it just will lack meaningful confidence (variation within a sample of 1 is enough). That's mathematics.
Could you elaborate on this? All the likelihood estimation techniques (and similar) I've ever seen have requirements on sample size, 1 is never enough - to say something has a probability of 0.5 ± 0.5 is to say you have no information at all - by definition probability is already in the range 0 ≤ p ≤ 1.
If it's happened once then you do know p > 0.