ubic.basecode.math.linearmodels

Class MeanVarianceEstimator

java.lang.Object

ubic.basecode.math.linearmodels.MeanVarianceEstimator

public class MeanVarianceEstimator
extends Object

Estimate mean-variance relationship and use this to compute weights for least squares fitting. R's limma.voom() Charity Law and Gordon Smyth. See Law et al. http://genomebiology.biomedcentral.com/articles/10.1186/gb-2014-15-2-r29

Running voom() on data matrices with NaNs is not currently supported.

Author:

ptan

Constructor Summary

Constructors

Constructor and Description
MeanVarianceEstimator(DesignMatrix designMatrix, DoubleMatrix<String,String> data, cern.colt.matrix.DoubleMatrix1D librarySize) Preferred interface if you want control over how the design is set up.
MeanVarianceEstimator(DesignMatrix designMatrix, cern.colt.matrix.DoubleMatrix2D data, cern.colt.matrix.DoubleMatrix1D librarySize) Executes voom() to calculate weights.
MeanVarianceEstimator(cern.colt.matrix.DoubleMatrix2D data) Generic method for calculating mean and variance, as a data diagnostic.

Method Summary

Modifier and Type	Method and Description
cern.colt.matrix.DoubleMatrix1D	getLibrarySize()
cern.colt.matrix.DoubleMatrix2D	getLoess()
cern.colt.matrix.DoubleMatrix2D	getMeanVariance()
cern.colt.matrix.DoubleMatrix2D	getNormalizedValue()
cern.colt.matrix.DoubleMatrix2D	getWeights()
protected cern.colt.matrix.DoubleMatrix1D	quarterRootVariance(LeastSquaresFit lsf)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

MeanVarianceEstimator

public MeanVarianceEstimator(DesignMatrix designMatrix,
                             DoubleMatrix<String,String> data,
                             cern.colt.matrix.DoubleMatrix1D librarySize)

Preferred interface if you want control over how the design is set up. Executes voom() to calculate weights.

Parameters:

designMatrix -

data - expected to be log2cpm, and already filtered

librarySize - library size (matrix column sum)

MeanVarianceEstimator

public MeanVarianceEstimator(DesignMatrix designMatrix,
                             cern.colt.matrix.DoubleMatrix2D data,
                             cern.colt.matrix.DoubleMatrix1D librarySize)

Executes voom() to calculate weights.

Parameters:

designMatrix -

data - a normalized count matrix

librarySize - library size (matrix column sum)

MeanVarianceEstimator

public MeanVarianceEstimator(cern.colt.matrix.DoubleMatrix2D data)

Generic method for calculating mean and variance, as a data diagnostic.

voom() is not executed and therefore no weights are calculated.

Parameters:

data - data to be processed

Method Detail

getLibrarySize

public cern.colt.matrix.DoubleMatrix1D getLibrarySize()

Returns:

total library size

getLoess

public cern.colt.matrix.DoubleMatrix2D getLoess()

Returns:

the loess fit of the mean-variance relationship

getMeanVariance

public cern.colt.matrix.DoubleMatrix2D getMeanVariance()

Returns:

the mean and variance of the normalized data, columns 0 and 1 respectively

getNormalizedValue

public cern.colt.matrix.DoubleMatrix2D getNormalizedValue()

Returns:

data, as supplied (should be log2cpm)

getWeights

public cern.colt.matrix.DoubleMatrix2D getWeights()

Returns:

inverse variance weights if voom was applied

quarterRootVariance

protected cern.colt.matrix.DoubleMatrix1D quarterRootVariance(LeastSquaresFit lsf)