3 ms·
Bayes rule is just: Pr(H0|Data) = Pr(H0)Pr(Data|H0)/sum(Pr(H[0:n])Pr(Data|H[0:n])) If the data indicates you lower the probability of any hypothesis 0:n, it m
by nonbel 8y ago
Bayes rule is just:
Pr(H0|Data) = Pr(H0)Pr(Data|H0)/sum(Pr(H[0:n])Pr(Data|H[0:n]))
If the data indicates you lower the probability of any hypothesis 0:n, it means the others all become more likely. There is no reason this has to be by a significant amount though.
- db48x 8y agoYea. My point was that with no way to fully enumerate the set of hypotheses, you can't do that sum. Thus the change is insignificant.
- nonbel 8y agoYou dont need to do the exact sum. Small values in the denominator have negligible effect and can be dropped. Also, "you cant do the sum, therefore its insignificant" makes no sense.