I'm trying to calculate the PCA (Principal Component Analysis) for my data using Minitab 17 statistical software, and when I tried to calculate the PCA for it, this error message (* ERROR * The covariance matrix is singular; computation cannot proceed) show up to me. So, if you please I need to know what does this message mean and how can I solve such a problem?!!
The problem occur in your analysis when the data in one row are exactly the same or a the data in another row. The same problem can also due to to the data in one row is a multiple of the data from another row. In mathematics, the term is called linear dependence. The algorithm could not find the eigenvalue since this is a "single" matrix.
Thank U sir, that was very helpful, and yes the message is totaly clear, it was just this part of it (The covariance matrix is singular) wasn't clear to me. Then U did clear it as U said the PCA doesn't work on extremely collinear variables. So I shall try what U have suggest, and see if does it work or not. thx alot.
The problem occur in your analysis when the data in one row are exactly the same or a the data in another row. The same problem can also due to to the data in one row is a multiple of the data from another row. In mathematics, the term is called linear dependence. The algorithm could not find the eigenvalue since this is a "single" matrix.
The mathematical core task of Principal Component Analysis is to perform a singular value decomposition of a given matrix (the covariance matrix). The linear dependences previous contributions speak about should only increase the number of zeros in the list of singular values and should pose no problem to a robust SVD algorithm. But real software is not always robust. Perhaps the developers considered it helpful to warn the user and let him know that his matrix is singular.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software