Re: principal components analysis

On Wed, 2012-11-07 at 21:46 -0500, Andrew Warshall wrote:
Gnumeric 1.10.17 seems to be having trouble doing principal components
analysis. On some (but not all) datasets, it produces a
reasonable-looking (and apparently symmetric) covariance matrix, but
the eigenvalues and eigenvectors are all "#NUM!", which (if I
understand your documentation correctly) means the matrix it was trying
to diagonalize (in this case the covariance matrix) was not in fact
symmetric. What am I missing?

There can also be rounding issues that may not allow us to calculate
eigenvectors and eigenvalues.

Perhaps you could provide a sample file where things fail that you think
should not, by filing an appropriate bug report at


Andreas Guelzow <aguelzow pyrshep ca>

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]