acPCoA_prediction: Perform AC-PCoA on new data points for classification and...

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

Description

Perform AC-PCoA on new data points for classification and prediction

Usage

acPCoA_prediction(
  training.dist,
  testing.dist,
  training.confounder,
  training.label,
  testing.label,
  nPC = 2,
  kernel = "linear",
  bandwidth = NULL,
  lambdas = seq(0, 5, 0.05),
  anov = T,
  perc = 0.05
)

Arguments

training.dist	the n by n data distance matrix, where n is the number of training samples.
testing.dist	the n by m data distance matrix, where n is the number of training samples and m is the number of testing samples.
training.confounder	the n by q confounder matrix of training samples, where n is the number of samples, q is the number of confounding factors.
training.label	Label of traning samples for classification. Optional.
testing.label	Label of testing samples for classification. Optional.
nPC	number of principal components to compute
kernel	the kernel to use: "linear", "gaussian".
bandwidth	bandwidth h for Gaussian kernel. Optional.
lambdas	the tuning parameter, non-negative.
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.

Value

principal components and the projected data of training and testing samples

v	the principal components, p by nPC matrix
Xv_training	the projected data of training samples, i.e. X times v
Xv_testing	the projected data of testing samples
eigenvalueX	eigenvalues for the PCs
varianceX	variance explained by the PCs
PredictingAccuracy	the accuracy of predicting labels of testing samples using Random Forest, if training and testing labels are provided.

Examples

## Not run: 
training.dist=data_mbqc_groupA$DistMat.BC[1:679,1:679]
testing.dist=data_mbqc_groupA$DistMat.BC[1:679,680:848]
training.confounder=data_mbqc_groupA$ConfounderMat[1:679,]
label=as.factor(data_mbqc_groupA$Specimen)
training.label=label[1:679]
testing.label=label[680:848]
result_prediction=acPCoA_prediction(training.dist,testing.dist,training.confounder,training.label,testing.label,nPC=2,kernel="linear",bandwidth=NULL,lambdas=seq(0, 5, 0.05),anov=T,perc=0.05)

## End(Not run)

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