ubic.basecode.math

Class MatrixStats

java.lang.Object

ubic.basecode.math.MatrixStats

public class MatrixStats
extends Object

Author:

pavlidis

Constructor Summary

Constructors

Constructor and Description
MatrixStats()

Method Summary

Modifier and Type	Method and Description
static <R,C> cern.colt.matrix.DoubleMatrix1D	colSums(DoubleMatrix<R,C> data) NaN values are omitted from calculations.
static <R,C> void	convertToLog2(DoubleMatrix<R,C> matrix, double base) Convert a log_b-transformed data set to log 2.
static <R,C> DoubleMatrix<R,C>	convertToLog2Cpm(DoubleMatrix<R,C> matrix, cern.colt.matrix.DoubleMatrix1D librarySize) Convert a count matrix to log2 counts per million.
static <R,C> DoubleMatrix<R,R>	correlationMatrix(DoubleMatrix<R,C> data) Compute the correlation matrix of the rows of a matrix.
static <R,C> SparseDoubleMatrix<R,R>	correlationMatrix(DoubleMatrix<R,C> data, double threshold)
static cern.colt.matrix.DoubleMatrix2D	cov2cor(cern.colt.matrix.DoubleMatrix2D cov) Scale a covariance matrix to the corresponding correlation matrix.
static <R,C> DoubleMatrix<R,C>	doubleStandardize(DoubleMatrix<R,C> matrix) Iteratively standardize the columns and rows of the matrix.
static <R,C> void	logTransform(DoubleMatrix<R,C> matrix) Log-transform the values in a matrix (log base 2).
static <R,C> double	max(DoubleMatrix<R,C> matrix) Compute the maximum value in the matrix.
static <R,C> double	min(DoubleMatrix<R,C> matrix) Find the minimum of the entire matrix.
static boolean[][]	nanStatusMatrix(double[][] data)
static <R,C> void	rbfNormalize(DoubleMatrix<R,C> matrixToNormalize, double sigma) Normalize a matrix in place to be a transition matrix.
static double[][]	selfSquaredMatrix(double[][] input)
static <R,C> DoubleMatrix<R,C>	standardize(DoubleMatrix<R,C> matrix) Scale the rows of the matrix; returns a new matrix.
static <R,C> void	unLogTransform(DoubleMatrix<R,C> matrix) Undo log2 transform.

Methods inherited from class java.lang.Object

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

Constructor Detail

MatrixStats

public MatrixStats()

Method Detail

colSums

public static <R,C> cern.colt.matrix.DoubleMatrix1D colSums(DoubleMatrix<R,C> data)

NaN values are omitted from calculations.

Type Parameters:

R -

C -

Parameters:

data -

Returns:

column sums

convertToLog2

public static <R,C> void convertToLog2(DoubleMatrix<R,C> matrix,
                                       double base)

Convert a log_b-transformed data set to log 2.

Parameters:

matrix -

base - the current base

convertToLog2Cpm

public static <R,C> DoubleMatrix<R,C> convertToLog2Cpm(DoubleMatrix<R,C> matrix,
                                                       cern.colt.matrix.DoubleMatrix1D librarySize)

Convert a count matrix to log2 counts per million. Equivalent to t(log2(t(counts+0.5)/(lib.size+1)*1e6)) in R. (NOTE: originally this did the operation in place)

Parameters:

matrix -

librarySize - if null, it will default to colSums(matrix).

Returns:

correlationMatrix

public static <R,C> DoubleMatrix<R,R> correlationMatrix(DoubleMatrix<R,C> data)

Compute the correlation matrix of the rows of a matrix.

Parameters:

data -

Returns:

a symmetric matrix that has the rows and columns set to be the names of the rows of the input.

correlationMatrix

public static <R,C> SparseDoubleMatrix<R,R> correlationMatrix(DoubleMatrix<R,C> data,
                                                              double threshold)

Parameters:

data - DenseDoubleMatrix2DNamed

threshold - only correlations with absolute values above this level are stored (others are Double.NaN)

Returns:

a sparse symmetric matrix that has the rows and columns set to be the names of the rows of the input. The diagonal is set to Double.NaN

doubleStandardize

public static <R,C> DoubleMatrix<R,C> doubleStandardize(DoubleMatrix<R,C> matrix)

Iteratively standardize the columns and rows of the matrix.

Parameters:

data -

logTransform

public static <R,C> void logTransform(DoubleMatrix<R,C> matrix)

Log-transform the values in a matrix (log base 2). Values that are less than or equal to zero are left as Double.NaN.

Parameters:

matrixToNormalize -

max

public static <R,C> double max(DoubleMatrix<R,C> matrix)

Compute the maximum value in the matrix.

Parameters:

matrix - DenseDoubleMatrix2DNamed

Returns:

the largest value in the matrix

min

public static <R,C> double min(DoubleMatrix<R,C> matrix)

Find the minimum of the entire matrix.

Parameters:

matrix - DenseDoubleMatrix2DNamed

Returns:

the smallest value in the matrix

nanStatusMatrix

public static boolean[][] nanStatusMatrix(double[][] data)

Parameters:

data -

Returns:

matrix indicating whether each value in the input matix is NaN.

rbfNormalize

public static <R,C> void rbfNormalize(DoubleMatrix<R,C> matrixToNormalize,
                                      double sigma)

Normalize a matrix in place to be a transition matrix. Assumes that values operate such that small values like p values represent closer distances, and the values are probabilities.

Each point is first transformed via v' = exp(-v/sigma). Then the values for each node's edges are adjusted to sum to 1.

Parameters:

matrixToNormalize -

sigma - a scaling factor for the input values.

selfSquaredMatrix

public static double[][] selfSquaredMatrix(double[][] input)

Parameters:

input - raw double 2-d matrix

Returns:

the element-by-element product (not matrix product) of the matrix.

standardize

public static <R,C> DoubleMatrix<R,C> standardize(DoubleMatrix<R,C> matrix)

Scale the rows of the matrix; returns a new matrix.

Type Parameters:

R -

C -

Parameters:

data -

Returns:

cov2cor

public static cern.colt.matrix.DoubleMatrix2D cov2cor(cern.colt.matrix.DoubleMatrix2D cov)

Scale a covariance matrix to the corresponding correlation matrix.

Parameters:

cov - a symmetric matrix of covariances

Returns:

unLogTransform

public static <R,C> void unLogTransform(DoubleMatrix<R,C> matrix)

Undo log2 transform.

Type Parameters:

R -

C -

Parameters:

matrix -