PCA Singular Value Decomposition (SVD) approach is faster and is recommended (set 'use.svd = TRUE')

[imren bayil]

I want to analysis pca

pc <- pca.xyz(xyz[,sel.ins$xyz]) I am getting this error

NOTE: In input xyz (MxN), N > 3000 and M < N
Singular Value Decomposition (SVD) approach is faster
and is recommended (set 'use.svd = TRUE') I dont why I am getting this error

[Xinqiu Yao]

Hi,

‌

This is not an error, just a suggestion. You should get the result anyway, but if it takes too long, you may think of following the suggestion.

[imren bayil]

ok.what is the suggestion to me ? how to use this suggestion as a code ?

[imren bayil]

pc <- pca.xyz(xyz[,sel.ins$xyz]) after this command,I am getting this warning

Error: cannot allocate vector of size 9.1 Gb

[Xinqiu Yao]

Try pc <- pca.xyz(xyz[, sel.ins$xyz], use.svd=TRUE)

[imren bayil]

I got this warning

pc <- pca.xyz(xyz[, sel.ins$xyz], use.svd=TRUE)
Warning message:
In pca.xyz(xyz[, sel.ins$xyz], use.svd = TRUE) : In input xyz (MxN), M < N:
Only 1015 eigenvalues and eigenvectors are returned!

[Xinqiu Yao]

The warning is fine. Your number of data points is less than the dimension of the degree of freedom. In PCA, you always get the number of eigenvalues/eigenvectors equal to the smaller number of the two.