random numbers with mean-correction

2 ビュー (過去 30 日間)
Fernando
Fernando 2013 年 6 月 9 日
Hi,
I would like to generate random numbers with follow a normal distribution (randn) but mean-corrected with the following condition: sum(randon numbers)=K.
Where K is a constant.
Can anyone help me?
Thanks a lot.

採用された回答

Walter Roberson
Walter Roberson 2013 年 6 月 9 日

その他の回答 (1 件)

Roger Stafford
Roger Stafford 2013 年 6 月 9 日
In preference to the method in fileexchange/9700, (even though I wrote it,) I would suggest a method that exhibits a normal distribution on the hyperplane defined by K.
Suppose each vector (sample) of numbers which must have a sum of K has n components and you wish to generate N such random vectors. Do this:
u = repmat(K/n,n,1);
V = null(u.');
P = bsxfun(@plus,u,V*randn(n-1,N));
Each of the N columns of P has n numbers which always sum to K. In the n-1 dimensional hyperplane defined by this requirement, the points do have a multivariate normal distribution.
  1 件のコメント
Fernando
Fernando 2013 年 6 月 9 日
Ok, thanks a lot. I didn't know that way of generate random numbers neither the functions null and bsxfun.

サインインしてコメントする。

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by