ubic.basecode.math.linearmodels

Class ModeratedTstat

java.lang.Object

ubic.basecode.math.linearmodels.ModeratedTstat

public class ModeratedTstat
extends Object

Implements methods described in

Smyth, G. K. (2004). Linear models and empirical Bayes methods for assessing differential expression in microarray experiments. Statistical Applications in Genetics and Molecular Biology Volume 3, Issue 1, Article 3.

R code snippets in comments are from squeezeVar.R in the limma source code.

Author:

paul

Field Summary

Fields

Modifier and Type	Field and Description
static double	TOOSMALL

Constructor Summary

Constructors

Constructor and Description
ModeratedTstat()

Method Summary

Modifier and Type	Method and Description
static void	ebayes(LeastSquaresFit fit) Does essentially the same thing as limma::ebayes
protected static double[]	fitFDist(cern.colt.matrix.DoubleMatrix1D vars, cern.colt.matrix.DoubleMatrix1D df1s)
protected static cern.colt.matrix.DoubleMatrix1D	squeezeVar(cern.colt.matrix.DoubleMatrix1D var, cern.colt.matrix.DoubleMatrix1D df, LeastSquaresFit fit) Ignoring robust and covariate for now
protected static cern.colt.matrix.DoubleMatrix1D	squeezeVariances(cern.colt.matrix.DoubleMatrix1D var, cern.colt.matrix.DoubleMatrix1D df, double[] fit)

Methods inherited from class java.lang.Object

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

Field Detail

TOOSMALL

public static final double TOOSMALL

Constructor Detail

ModeratedTstat

public ModeratedTstat()

Method Detail

ebayes

public static void ebayes(LeastSquaresFit fit)

Does essentially the same thing as limma::ebayes

Parameters:

fit - which will be modified

fitFDist

protected static double[] fitFDist(cern.colt.matrix.DoubleMatrix1D vars,
                                   cern.colt.matrix.DoubleMatrix1D df1s)

squeezeVar

protected static cern.colt.matrix.DoubleMatrix1D squeezeVar(cern.colt.matrix.DoubleMatrix1D var,
                                                            cern.colt.matrix.DoubleMatrix1D df,
                                                            LeastSquaresFit fit)

Ignoring robust and covariate for now

Parameters:

var - initial values of estimated residual variance = sigma^2 = rssq/rdof; this will be moderated

df - vector of degrees of freedom

fit - will be updated with new info; call fit.summarize() to get updated pvalues etc.

Returns:

varPost for testing mostly

squeezeVariances

protected static cern.colt.matrix.DoubleMatrix1D squeezeVariances(cern.colt.matrix.DoubleMatrix1D var,
                                                                  cern.colt.matrix.DoubleMatrix1D df,
                                                                  double[] fit)

Parameters:

var - vector of estimated residual variances from original model fit

df - vector of dfs

fit - result of fitFDist() (s2.prior and dfPrior)

Returns:

vector of squeezed variances (varPost or s2.post)