PPTrank: Determine the best low rank using Projection based...

In YunhuiQi/TestPMD: Post Selection Inference on Results from Penalized Matrix Decomposition (SCCA only)

Determine the best low rank using Projection based Permutation Test (PPT)

Description

This function sequentially performs Projection based Permutation Test on singular values of cross-covariance matrix between data matrices X and Y, thus determine the best low rank.

Usage

PPTrank(
  X,
  Y,
  max,
  penalty = c("Fixed", "CV"),
  rho_x = NULL,
  rho_y = NULL,
  permutation_no,
  alpha = 0.05
)

Arguments

X	Data matrix, each row is one sample, each column is one feature.
Y	Data matrix, each row is one sample, each column is one feature.
max	The largest index that will be considered.
penalty	"Fixed" or "CV": how to choose the penalty parameter, using fixed value or through cross validation.
rho_x	Penalty parameter used for PMD estimation of data X. If penalty = "Fixed", rho should be a single value, if penalty = "CV", rho_x should be a vector containing candidate penalty parameters for cross validation.
rho_y	Penalty parameter used for PMD estimation of data Y. If penalty = "Fixed", rho should be a single value, if penalty = "CV", rho_y should be a vector containing candidate penalty parameters for cross validation.
permutation_no	Integer: number of permutations for each test.
alpha	Significance level, 0.05 by default

Value

A list containing best low rank and p-values

Examples

library(TestPMD)
data("covid")
PPTrank(X = covid$metabolite, Y = covid$protein, max = 2, penalty = "Fixed",
 rho_x = 0.9, rho_y = 0.9, permutation_no = 100, alpha = 0.1)

YunhuiQi/TestPMD documentation built on May 5, 2022, 8:23 p.m.