understending PCA final components

3 次查看（过去 30 天）

Dany 2015-1-14

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/169898-understending-pca-final-components

评论： Dany 2015-1-14

Hi, i have a question about PCA results. if i am analysing tidal data record using PCA i am getting first two PC's that look the same, the only difference is that they have a slight phase shift. when i plot them with a scatter plot, each one of them on a different axis i get almost a perfect circle. the longer my data record is, the more this circle becomes perfect.

can someone explain what does it mean? does it mean anything at all? if this is not the place for this question maybe you can redirect me to a different forum.

thank you. Dany

2 个评论
显示无隐藏无

Adam 2015-1-14

I'm not an expert on PCA, but if you get two components almost equally dominant (by score) giving you close to a circle that suggests your data is not dominantly correlated in any particular direction and is fairly random with respect to those two dominant components.

The components should always be orthogonal to each other though which would tend to give you a circle if they are close to equal.

Dany 2015-1-14