fabatch {bapred}

Batch effect adjustment using FAbatch

Description

Performs batch effect adjustment using the FAbatch-method described in Hornung et al. (2016) and additionally returns information necessary for addon batch effect adjustment with FAbatch.

Usage

fabatch(x, y, batch, nbf = NULL, minerr = 1e-06, 
  probcrossbatch = TRUE, maxiter = 100, maxnbf = 12)

Arguments

x

matrix. The covariate matrix. Observations in rows, variables in columns.

y

factor. Binary target variable. Has to have two factor levels, where each of them correponds to one of the two classes of the target variable.

batch

factor. Batch variable. Each factor level (or 'category') corresponds to one of the batches. For example, if there are four batches, this variable would have four factor levels and observations with the same factor level would belong to the same batch.

nbf

integer. Number of factors to estimate in all batches. If not given the number of factors is estimated automatically for each batch. Recommended to leave unspecified.

minerr

numeric. Maximal mean quadratic deviations between the estimated residual variances from two consecutive iterations. The iteration stops when this value is undercut.

probcrossbatch

logical. Default is TRUE. If TRUE the preliminary probabilities are estimated through leave-one-batch-out cross-validation. If set to FALSE ordinary cross-validation is used for estimating the preliminary probabilities. This might result in an artificially increased class signal in comparison to that in the data in independent batches. Is automatically set to FALSE, when only one batch is present in the training data.

maxiter

integer. Maximal number of iterations in the estimation of the latent factors by Maximum Likelihood.

maxnbf

integer. Maximal number of factors if nbf is not given. Default is the largest integer smaller than half the number of observations in the corresponding batch.

Value

fabatch returns an object of class fabatch. An object of class "fabatch" is a list containing the following components:

xadj

matrix of adjusted (training) data

m1

means of the standardized variables in class '1'

m2

means of the standardized variables in class '2'

b0

intercept out of the L2-penalized logistic regression performed for estimation of the class probabilities

b

variable coefficients out of the L2-penalized logistic regression performed for estimation of the class probabilities

pooledsds

vector containing the pooled standard deviations of the variables

meanoverall

vector containing the variable means

minerr

maximal mean quadratic deviations between the estimated residual variances from two consecutive iterations

nbfinput

user-specified number of latent factors nbf in all batches. NULL if nbf was not specified.

badvariables

indices of those variables which are constant in at least one batch

nbatches

number of batches

batch

batch variable

nbfvec

vector containing the numbers of factors in the individual batches

Author(s)

Roman Hornung

References

Hornung, R., Boulesteix, A.-L., Causeur, D. (2016). Combining location-and-scale batch effect adjustment with data cleaning by latent factor adjustment. BMC Bioinformatics 17:27, <doi: 10.1186/s12859-015-0870-z>.

Examples

data(autism)

# Random subset of 150 variables:
set.seed(1234)
Xsub <- X[,sample(1:ncol(X), size=150)]

# In cases of batches with more than 20 observations
# select 20 observations at random:
subinds <- unlist(sapply(1:length(levels(batch)), function(x) {
  indbatch <- which(batch==x)
  if(length(indbatch) > 20)
    indbatch <- sort(sample(indbatch, size=20))
  indbatch
}))
Xsub <- Xsub[subinds,]
batchsub <- batch[subinds]
ysub <- y[subinds]



fabatch(x=Xsub, y=ysub, batch=batchsub)