用正太分布的近似方法计算二项分布的概率的置信区间如果上限超过1下线低于0

用正太分布的近似方法计算二项分布的概率的置信区间如果上限超过1下线低于0

所有跟贴·加跟贴·新语丝读书论坛

送交者: HunHunSheng 于 2008-02-06, 18:47:19:

Do I need to set upper limit at 1 and lower limit at zero?

p=prob(yes)

Upper=p+1.96*sqrt(p(1-p)/sampel size)
Lower=p-1.96*sqrt(p(1-p)/sampel size)

p range from 0 to 1 but Upeer and lower can exceed this range

sorry, usually for extremely small p, set the lower limit to 0, vice versa (无内容) - xj (0 bytes) 2008-02-06, 19:00:56 (203812)
usually won't. The largest SE for bin(n,p) is when p=0.5 - xj (55 bytes) 2008-02-06, 18:55:22 (203809)
- 从理论上讲alpha可以取得很小很小置信区间会越来越宽 (无内容) - HunHunSheng (0 bytes) 2008-02-06, 20:17:53 (203829)
  - there is a rule of thumb like np>5 to approximate normal dist (无内容) - xj (0 bytes) 2008-02-06, 22:30:36 (203834)
    - 你丫不要胡说八道即使大于5置信区间也有可能两头超出百分之百和0 (无内容) - HunHunSheng (0 bytes) 2008-02-07, 08:05:06 (203858)