error in multilevel PCA

Issue #119 new
Kim-Anh Le Cao repo owner created an issue

Email from Arno V

When conducting the multilevel PCA, it works on some parts of the data, but other parts give following error: 'Error in ph.new/drop(sqrt(crossprod(ph.new))) : non-conformable arrays'. When performing multilevel PLS-DA on the same data, it does not give an error, just a warning: 'In cor(Ak, variates.Ak) : the standard deviation is zero'. Is there a way to solve this error in PCA and is it related to the warning in PLS-DA?
Thanks in advance!

Comments (1)

  1. Florian Rohart

    Looks like there's NA in the data "ph.new/drop(sqrt(crossprod(ph.new))" comes from the nipals algorithm. I'd say there's probably parameters with no variance, and PCA doesn't remove any parameters, so maybe removing some parameters as a pre-processing will help.

    I can't say much more at this point. Data would help

  2. Log in to comment