T O P I C R E V I E W |
Cnb |
Posted - 09/23/2015 : 08:29:19 AM Origin Ver. and Service Release (Select Help-->About Origin): OriginPro 2015 (64-bit) Sr2 Operating System: Windows
Hoping someone can help clarify an issue I'm having with PCA in Origin.
I have standardised some data and undertaken PCA on it (covariance matrix). The PC Scores should be equal to the sum of the standardised data multiplied by the respective Eigenvectors (Coefficients of PC1, PC2 etc.). This is not the case in my example and I cannot see how Origin has arrived at the results.
In a simple test I have run with two columns of non-standardised data, I can arrive at the PC Scores output by Origin (covariance matrix) only if I subtract the mean from each observation (causing mean of adjusted observations to be 0) before multiplying it by the respective Eigenvector and summing. Alternatively, I can subtract the mean from the original data before undertaking PCA and get the correct result too. I think this makes sense because I am told the mean must be 0 for PCA to work correctly.
However, with my standardised data obviously the mean is also 0 yet I cannot get to the PC scores output by Origin using the coefficients and input standardised data. Has anyone else noticed this and know what's going on?
|
1 L A T E S T R E P L I E S (Newest First) |
Sam Fang |
Posted - 09/24/2015 : 02:37:41 AM I can't reproduce your problem. For covariance matrix, PC Scores are the multiplication of column mean-subtracted input data matrix and eigenvectors.
Note that you should uncheck Standardize Scores option.
If you still have the question, you can send your result to us. And we can check it.
To send files to us, click Send File to Tech support button on the top right of the forum.
Thanks.
Sam OriginLab Technical Services |
|
|