她的意思是方差太小所以最大值很大的可能性不大

所有跟贴·加跟贴·新语丝读书论坛

送交者: HunHunSheng 于 2007-09-27, 14:40:41:

回答: 我把短江的推理再加强一把：中原不仅人均智商高于欧美，大牛由 chouqilozi 于 2007-09-27, 14:37:23:

If x fowllows a normal distrubutiuon with u as meand and sigma as atd, what is the distribution of y=max(x)?

And what is the mean of Y?

所有跟贴:

normal distribution has max? better to argue it with chebychev - mangolasi (55 bytes) 2007-09-27, 14:47:06 (173326)
- 我说的不严谨. 我说得是max(x1,x2,...xn) while xi follow normal dist - HunHunSheng (28 bytes) 2007-09-27, 14:52:59 (173333)
  - wait..the distribution of max(x1,...xn) and max(y1,...ym) also - mangolasi (490 bytes) 2007-09-27, 15:23:29 (173344)
    - 首先这是理论现实中没有真正的正态分布 - HunHunSheng (83 bytes) 2007-09-27, 16:59:26 (173394)
      - The problem is, starting from longer right tail for the european+ - mangolasi (308 bytes) 2007-09-27, 17:27:10 (173407)
      - The only assumption is iid here (not a weak one I agree). No normal - mangolasi (297 bytes) 2007-09-27, 17:21:59 (173406)
    - 你们这上下在讨论啥呢，一会cdf,一会pdf的 - xj (10 bytes) 2007-09-27, 16:06:49 (173369)
      - How to convert an cdf file to pdf file? I guess (无内容) - blackbox (0 bytes) 2007-09-27, 16:54:44 (173390)
    - correct, so we are worse than we thought! :) (无内容) - chouqilozi (0 bytes) 2007-09-27, 15:26:47 (173350)
      - No. P(max(x_i)>z)=1-P(max(x_i)<z)=1-(G1(X<z))^n - 湘女 (53 bytes) 2007-09-27, 15:55:44 (173368)
        
        He means for large, genius level z, G1(X<=z) is so large (means - mangolasi (869 bytes) 2007-09-27, 16:12:08 (173371)
        
        明白了，谢谢！结论是挺吓人的。但这应是教育的结果，而不是人种的差别。 (无内容) - 湘女 (0 bytes) 2007-09-27, 16:29:38 (173378)
        
        天才! 华人教授坐校巴想老半天也没看出个四五六你一眼就明白了,佩服 (无内容) - blackbox (0 bytes) 2007-09-27, 17:00:17 (173395)
      - so you mean for some large, genius level z, G1(X<=z) is so large - mangolasi (125 bytes) 2007-09-27, 15:43:24 (173358)
  - yup...you can get the distribution of this...use permutaion... (无内容) - mangolasi (0 bytes) 2007-09-27, 15:00:08 (173338)
    - extremes of normal dist usually follows weibull class - xj (42 bytes) 2007-09-27, 15:25:34 (173348)
    - pdf should be nF^(n-1)(x)f(x) - HunHunSheng (31 bytes) 2007-09-27, 15:20:09 (173342)
      - why care about pdf? Better to use cdf if you want to check the - mangolasi (31 bytes) 2007-09-27, 15:24:57 (173347)
        
        never heard of extreme value theory? - xj (44 bytes) 2007-09-27, 15:26:36 (173349)

加跟贴