Covariance matrix not always positive define

2 次查看（过去 30 天）

Gina Carts 2019-3-21

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/451394-covariance-matrix-not-always-positive-define

评论： Adam 2019-3-21

I have used the cov Matlab function to calculate the covariance matrix.

I calculate the covariance matrix inside a loop and this stop by giving me the error "Error using iwishrnd (line 41) Covariance matrix must be symmetric and positive definite."

I printed the covariance matrix to see how it looks like and basically after the second calculation of the covariance the program stopped. Is there any way to control the covariance matrix in order to be always positive define?

covar =
    1.1658    0.6780   -0.0203   -0.1593   -0.0339
    0.6780    8.0017   -0.3713    0.0057   -0.3568
   -0.0203   -0.3713    1.0419   -0.1423   -0.0006
   -0.1593    0.0057   -0.1423    1.1528    0.0454
   -0.0339   -0.3568   -0.0006    0.0454    0.8498
covar =
    1.3151    0.1604    0.1350         0         0
    0.1604    1.5119   -0.2529         0         0
    0.1350   -0.2529    0.7174         0         0
         0         0         0         0         0
         0         0         0         0         0

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Adam 2019-3-21

You can often get away with adding a small value to the leading diagonal, but as with anything that changes your data to avoid issues like dividing by zero, it needs to be carefully considered and may not always work.

请先登录，再进行评论。

请先登录，再回答此问题。