Calculate the principal components of an array of correlation or
covariance matrices.

# S3 method for array
pca(x, use.svd = TRUE, rm.gaps=TRUE, ...)

## Arguments

x |
an array of matrices, e.g. correlation or covariance
matrices as obtained from functions `dccm` or `enma2covs` . |

use.svd |
logical, if TRUE singular value decomposition (SVD) is
called instead of eigenvalue decomposition. |

rm.gaps |
logical, if TRUE gap cells (with missing
coordinate data in any input matrix) are removed before
calculation. This is equivalent to removing NA cells from x. |

... |
. |

## Details

This function performs PCA of symmetric matrices, such as distance
matrices from an ensemble of crystallographic structures, residue-residue
cross-correlations or covariance matrices derived from ensemble NMA
or MD simulation replicates, and so on. The ‘upper triangular’
region of the matrix is regarded as a long vector of random variables.
The function returns M eigenvalues and eigenvectors with each eigenvector
having the dimension N(N-1)/2, where M is the number of matrices and N
the number of rows/columns of matrices.

## Value

Returns a list with components equivalent to the output from
`pca.xyz`

.

## References

Grant, B.J. et al. (2006) *Bioinformatics* **22**, 2695--2696.

## Author

Xin-Qiu Yao, Lars Skjaerven

## See also