Hi,
I’m trying to use TPrincipal to do some PCA on my data. I performed it on the following dummy data:
x y z
0.11547 1.8 -1.458
0.11547 1.8 -1.35
0.11547 1.8 -1.242
0.11547 1.8 -1.134
0.11547 1.8 -1.026
0.11547 1.8 -0.918
0.11547 1.8 -0.81
The x and the y values are constant, so this should just represent a straight line perpedicular to the z=0 plane. The sigma vector I get from TPrincipal is:
| 1 |
0 |2.1817e-17
1 |1.2911e-16
2 |0.216
This makes sense, the first two values are near zero as they should be.
However, this is what the co-variance matrix looks like:
| 0 | 1 | 2 |
0 | 0.3333 0.2466 -0.3256
1 | 0.2466 0.3333 -0.2662
2 | -0.3256 -0.2662 0.3333
As I understand, the diagonal elements of this should represent the variance of x,y and z variables which we know are not all equal to .3333, infact, none of them are.
Can someone help me make sense out of this ?
Thanks