PCA eigenvalues

mothur bugs
1

Hi Scott,

The following might just be my misunderstanding. I assume that the loadings file output from the pca command lists the eigenvalues adjusted so that they sum to 100. I’m finding that when I run principal component analysis in MatLab that the adjusted eigenvalues differ by as much as 7 or 8 (w.r.t. 100).

Cheers,
Stephen

2

In fact, there may not really be an issue at all. I obtained the differences when I truncated data sets to small matrices, e.g. 3 by 4. The result are approximately equal for 6 by 9 matrices or larger, so it is never actually a problem, as our datasets typically run into hundreds of OTUs.