m****a 发帖数: 604 | 1 Binominal 分布,每次trial要么出现1,要么0的题.
假定预计出现1的事前概率, denoted by p.现在做了n次实验,出现了m次1.请问,出现1
的事后概率, denoted by q,大于事前慨率的可能性是多少? | s********0 发帖数: 3644 | | J*******g 发帖数: 267 | 3
Binominal 分布,每次trial要么出现1,要么0的题.
假定预计出现1的事前概率, denoted by p.现在做了n次实验,出现了m次1.请问,出现1
~~~~~~~~~~~~~~~~~~~~~~~~~~with this
statement, your question is confusing
的事后概率, denoted by q,大于事前慨率的可能性是多少?
【在 m****a 的大作中提到】 : Binominal 分布,每次trial要么出现1,要么0的题. : 假定预计出现1的事前概率, denoted by p.现在做了n次实验,出现了m次1.请问,出现1 : 的事后概率, denoted by q,大于事前慨率的可能性是多少?
| t********s 发帖数: 54 | 4 seraphic80,可以讲讲你打算怎么用Bayesian method吗?
我做了如下计算,但是觉得不对。。。
假定A表示出现1的事件, P(A) = p
B表示出现0的事件,P(B) = 1-p
F表示n次trial出现m次1的事件,那么P(F)=C(n,m)*p^m*(1-p)^(n-m)
那么时候概率q应该是
P(A|F)=P(AF)/P(F)
但是P(AF) = P(F|A)*P(A),那么P(F|A)又是什么呢?如果A是指任一次trial,那P(F|A
)=C(n,m)*p^m*(1-p)^(n-m),那这样P(A|F)不是就等于P(A)了?
是不是不应该这么定义事件? | t********s 发帖数: 54 | 5 Looks like this might explain it: http://en.wikipedia.org/wiki/Checking_whether_a_coin_is_fair
Let F1 be the variable representing number of 1's during n trials, F0 be the
variable representing number of 0's during n trials. then the posterior
probability of a trial shows up with 1 is (assuming p is uniformly
distributed between 0 and 1):
q = Pr(p|F1=m, F0=n-m) =(n+1)!/(m!(n-m)!) * p^m * (1-p)^(n-m)
so we are looking at Pr(q>p). We can probably solve
q-p>0
to obtain a range of p, which would re |
|