com.xinapse.dynamic
Class AutoCorrelationEstimate

java.lang.Object
  com.xinapse.dynamic.AutoCorrelationEstimate

public class AutoCorrelationEstimate
extends java.lang.Object

Class to estimate the auto-correlation of a time-series by examining the residuals of a GLM fit. The auto-correlation is estimated pixel-by-pixel and then a spatial smoothing of the auto-correlation values is applied. The method follows that in: "Temporal Autocorrelation in Univariate Linear Modeling of FMRI Data" Moolrich MW et al NeuroImage 14: 1370-1386 (2001).

Constructor Summary

AutoCorrelationEstimate(float[][] residuals, int nCols, int nRows, int nSlices, float pixelXSize, float pixelYSize, float pixelZSize, MonitorWorker worker, boolean verbose)
Estimate the auto-correlation for the supplied residuals.

Method Summary

float[]

getVarContrasts(java.util.List<com.xinapse.dynamic.GLMContrastVector> contrastVectors, int col, int row, int slice, float[][] moorePenrosePsuedoInverse, float[][] moorePenrosePsuedoInverseTranspose, float[][] residualFormingMatrix, double sumSqResiduals) Get the estimated variances of a set of contrast of Beta (fitted parameter) estimates for a single pixel using this auto-correlation estimate.

Methods inherited from class java.lang.Object

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

Constructor Detail

AutoCorrelationEstimate

public AutoCorrelationEstimate(float[][] residuals,
                               int nCols,
                               int nRows,
                               int nSlices,
                               float pixelXSize,
                               float pixelYSize,
                               float pixelZSize,
                               MonitorWorker worker,
                               boolean verbose)
                        throws CancelledException

Estimate the auto-correlation for the supplied residuals. The auto-correlation can be used for variance correction for estimating the variance of the fitted Beta parameters in a GLM.

Parameters:

residuals - the residuals after fitting using a GLM. The first array index refers to the pixel index, and the second array index refers to the time point.

nCols - number of image columns.

nRows - number of image rows.

nSlices - number of image slices.

pixelXSize - the horizontal pixel size.

pixelYSize - the vertical pixel size.

pixelZSize - the slice spacing.

worker - if non-null, the MonitorWorker that is performing the estimate.

verbose - if verbose reporting to System.out is turned on.

Throws:

CancelledException - if calculation of the auto-correlation is cancelled by the user.

Method Detail

getVarContrasts

public float[] getVarContrasts(java.util.List<com.xinapse.dynamic.GLMContrastVector> contrastVectors,
                               int col,
                               int row,
                               int slice,
                               float[][] moorePenrosePsuedoInverse,
                               float[][] moorePenrosePsuedoInverseTranspose,
                               float[][] residualFormingMatrix,
                               double sumSqResiduals)

Get the estimated variances of a set of contrast of Beta (fitted parameter) estimates for a single pixel using this auto-correlation estimate.

Parameters:

contrastVectors - the set of contrast vectors for which to assess the variances.

col - the column number for the pixel.

row - the row number for the pixel.

slice - the slice umber for the pixel.

moorePenrosePsuedoInverse - the Moore-Penrose pseudo inverse for the GLM design matrix.

moorePenrosePsuedoInverseTranspose - the transpose of the Moore-Penrose pseudo inverse for the GLM design matrix.

residualFormingMatrix - the residual-forming matrix.

sumSqResiduals - the sum of squares of the residuals.

Returns:

a vector of variance estimates for the contrasts in the fitted Beta parameters of the GLM.