acPCAtuneLambda: Tune the lambda parameter in function acPCA

In YuWang28/acPCoA: Adjustment for Confounding Factors using Principal Coordinate Analysis

Description

Tune the lambda parameter in function acPCA

Usage

acPCAtuneLambda(
  X,
  Y,
  nPC,
  lambdas,
  centerX = T,
  centerY = T,
  scaleX = F,
  scaleY = F,
  kernel = c("linear", "gaussian"),
  bandwidth = NULL,
  anov = T,
  perc = 0.05,
  quiet = F
)

Arguments

X	the n by p data matrix, where n is the number of samples, p is the number of variables. Missing values in X should be labeled as NA. If a whole sample in X is missing, it should be removed.
Y	the n by q confounder matrix, where n is the number of samples, q is the number of confounding factors. Missing values in Y should be labeled as NA.
nPC	number of principal components to compute.
lambdas	a vector with the tuning parameters, non-negative values. If 0 is not in lambdas, it will be added to lambdas.
centerX	center the columns in X. Default is True.
centerY	center the columns in Y. Default is True.
scaleX	scale the columns in X to unit standard deviation. Default is False.
scaleY	scale the columns in Y to unit standard deviation. Default is False.
kernel	the kernel to use: "linear", "gaussian".
bandwidth	bandwidth h for Gaussian kernel. Optional.
anov	True or False. Whether the penalty term has the between groups sum of squares interpretation. Default is True.
perc	the best lambda is defined to be the smallest lambda with R(lambda)<=perc (if anov=T), or R(lambda)<=perc*R(lambda=0) (if anov=F) in the nPC principal components.
quiet	True or False. Output the progress of the program. Default is False.

Value

Results for tuning lambda

ratio	R(lambda): a vector with the ratios. Same length as lambdas
best_lambda	the best lambda after cross-validation
...	Input parameters for the function

YuWang28/acPCoA documentation built on Dec. 18, 2021, 8:20 p.m.