Package ubic.gemma.core.analysis.expression.coexpression.links

Class PearsonMetrics

java.lang.Object

ubic.gemma.core.analysis.expression.coexpression.links.AbstractMatrixRowPairAnalysis

ubic.gemma.core.analysis.expression.coexpression.links.PearsonMetrics

All Implemented Interfaces:

MatrixRowPairAnalysis

Direct Known Subclasses:

SpearmanMetrics

public class PearsonMetrics
extends AbstractMatrixRowPairAnalysis

A correlation analysis for a given data set, designed for selection of values based on criteria set by the user. On the first pass over the data, a histogram is filled in to hold the distribution of the values found. You can set criteria to have the correlations actually stored in a (sparse) matrix. This can take a lot of memory if you store everything! The correlation is only calculated if it isn't stored in the matrix, and values can be tested against a threshold. This class is used in reality by one pass over the data to fill in the histogram. This is used to help select a threshold. A second pass over the data is used to select correlations that meet the criteria. Probes that do not map to genes are not used.

Author:

Paul Pavlidis

Field Summary

Fields inherited from class ubic.gemma.core.analysis.expression.coexpression.links.AbstractMatrixRowPairAnalysis

HARD_LIMIT_MIN_NUM_USED

Fields inherited from interface ubic.gemma.core.analysis.expression.coexpression.links.MatrixRowPairAnalysis

NUM_BINS

Constructor Summary

Constructors

Constructor	Description
PearsonMetrics(ExpressionDataDoubleMatrix dataMatrix)
PearsonMetrics(ExpressionDataDoubleMatrix dataMatrix, double tmts)

Method Summary

Modifier and Type	Method	Description
void	calculateMetrics()	Calculate the linear correlation matrix of a matrix, allowing missing values.
protected double	correlFast(double[] ival, double[] jval, int i, int j)
QuantitationType	getMetricType()
protected void	rowStatistics()	Calculate mean and sumsqsqrt for each row

Methods inherited from class ubic.gemma.core.analysis.expression.coexpression.links.AbstractMatrixRowPairAnalysis

getCrossHybridizationRejections, getHistogramArrayList, getKeepers, getMatrix, getNumUniqueGenes, getProbeForRow, getScoreInBin, isUsePvalueThreshold, kurtosis, nullMatrix, numCached, setDuplicateMap, setLowerTailThreshold, setMinNumpresent, setOmitNegativeCorrelationLinks, setPValueThreshold, setUpperTailThreshold, setUseAbsoluteValue, setUsePvalueThreshold, size, toString

Methods inherited from class java.lang.Object

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

Constructor Details

PearsonMetrics

public PearsonMetrics(ExpressionDataDoubleMatrix dataMatrix)

PearsonMetrics

public PearsonMetrics(ExpressionDataDoubleMatrix dataMatrix,
 double tmts)

Parameters:

dataMatrix - DenseDoubleMatrix2DNamed

tmts - Values of the correlation that are deemed too small to store in the matrix. Setting this as high as possible can greatly reduce memory requirements, but can slow things down.

Method Details

calculateMetrics

public void calculateMetrics()

Calculate the linear correlation matrix of a matrix, allowing missing values. If there are no missing values, this calls PearsonFast.

getMetricType

public QuantitationType getMetricType()

rowStatistics

protected void rowStatistics()

Calculate mean and sumsqsqrt for each row

Specified by:

rowStatistics in class AbstractMatrixRowPairAnalysis

correlFast

protected double correlFast(double[] ival,
 double[] jval,
 int i,
 int j)