Obviously wrong answers for beta cdf

2 次查看（过去 30 天）

John Hey 2016-8-18

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/300020-obviously-wrong-answers-for-beta-cdf

编辑： John D'Errico 2016-8-18

Dear All,

I am using cdf('beta',x,alp,bet) to compute the cdf of a beta distribution but when alp and bet are less than 1 (where the pdf is u-shaped) it gives obviously wrong answers. For example cdf('beta',0.5,0.0003,.2277) it gives the answer 0.9987.

I do hope you can help. Thanks.

John Hey

2 个评论
显示无隐藏无

Torsten 2016-8-18

Mathematica gives the same result ...

http://www.wolframalpha.com/input/?i=Plot%5BTable%5BCDF%5BBetaDistribution%5B0.0003,0.2277%5D,x%5D%5D%2F%2FEvaluate,%7Bx,0,1%7D,Filling-%3EAxis%5D

Best wishes

Torsten.

John D'Errico 2016-8-18

编辑：John D'Errico 2016-8-18

And why do you think that is wrong? Perhaps the obviously wrong thing here is your knowledge of the expected result? That seems not unreasonable as a result. REALLY tiny values of the exponents will give nasty looking beta distributions, with a very heavy tail when only one parameter is small. So you should explain why it is that you think this result is wrong.

请先登录，再进行评论。

请先登录，再回答此问题。