transform_pcs: Create direct SNP weights from a reference

In danjlawson/pcapred: Produce PC Eigenvector Predictions From Plink Input Data And Loadings

Description

PCA can be represented in an a number of different ways, provided that the structure $$Y = U D V^T$$ flashpca uses this canonical form but the weights are out by a factor √{2} due to diploid genomes. Plink-like loadings instead scale U by the square root of the eigenvalues.

Usage

transform_pcs(snpweights, pcvals, mode)

Arguments

snpweights	A weighting matrix of dimension L by K
pcvals	Eigenvalues from the original PCA
mode	"plink" pr "flashpca"

Value

a canonical scaled version of the snpweights matrix

See Also

transform.plinkref

danjlawson/pcapred documentation built on Oct. 7, 2021, 3:47 p.m.