Tidy a gt_pca object

This summarizes information about the components of a gt_pca from the tidypopgen package. The parameter matrix determines which element is returned. Column names of the tidied output match those returned by broom::tidy.prcomp, the tidier for the standard PCA objects returned by stats::prcomp.

# S3 method for class 'gt_pca'
tidy(x, matrix = "eigenvalues", ...)

Arguments

x

A gt_pca object returned by one of the gt_pca_* functions.

matrix

Character specifying which component of the PCA should be tidied.

  • "samples", "scores", or "x": returns information about the map from the original space into principle components space (this is equivalent to product of u and d).

  • "v", "rotation", "loadings" or "variables": returns information about the map from principle components space back into the original space.

  • "d", "eigenvalues" or "pcs": returns information about the eigenvalues.

...

Not used. Needed to match generic signature only.

Value

A tibble::tibble with columns depending on the component of PCA being tidied.

If "scores" each row in the tidied output corresponds to the original data in PCA space. The columns are:

row

ID of the original observation (i.e. rowname from original data).

PC

Integer indicating a principal component.

value

The score of the observation for that particular principal component. That is, the location of the observation in PCA space.

If matrix is "loadings", each row in the tidied output corresponds to information about the principle components in the original space. The columns are:

row

The variable labels (colnames) of the data set on which PCA was performed.

PC

An integer vector indicating the principal component.

value

The value of the eigenvector (axis score) on the indicated principal component.

If "eigenvalues", the columns are:

PC

An integer vector indicating the principal component.

std.dev

Standard deviation (i.e. sqrt(eig/(n-1))) explained by this PC (for compatibility with prcomp.

cumulative

Cumulative variation explained by principal components up to this component (note that this is NOT phrased as a percentage of total variance, since many methods only estimate a truncated SVD.