PCA using NIPALS rather then SVD

[ombra...@gmail.com]

Hey

AS I understand it, the PrincipalComponentAnalysis class is based on SVD decomposition, and not on NIPALS like PartialLeastSquaresAnalysis?

I want to replace CAMEO's Unscrambler, in an educational course, with software based on the Accord.Net framework.

One important factor in multivariate analysis is the ability to handle missing data.

With respect to the implementation of PCA and PLS, how will the Accord.Net framework handle missing data? Will PLS with NIPALS be able to do this? AS I understand the implementation of PCA based on SVD, it has problems with exactly these missing data. Is there any option (or plan to include one) to compute PCA with the NIPALS algorithm? If there is already an implementation of this in the PLS class, maybe its not to complicated to add an option to the PCA class to do this also?