shuffle (version 1.0.1)

shuffle-package: The Shuffle Estimator for Explainable Variance

Description

This package implements the algorithms underlying the shuffle estimators, variance estimators for one-way analysis of variance designs. The estimators can overcome correlated noise by recomputing the mean-square-between statistics on a permuted version of the data. The permutations should preserve the noise covariance matrix, but a parametric model for the noise covariance is not necessary. For more details see Benjamini and Yu, and here http://statweb.stanford.edu/~yuvalben. Two functions implement the important stages of estimation: prepareShuffle(design_vec, premutation), which preprocess the design and computes the normalization constant for a given permutation. estimateShuffle(response_vec, prepare), which estimates variances and effect sizes for a specific data vector.

Details

Package:

shuffle

Type:

Package

Version:

1.0.1

Date:

2013-4-24

License:

GPL (>= 2)

References

Benjamini and Yu (2013), "The shuffle estimator for explainable variance in fMRI experiments", Annals of Applied Statistics 7 (4) http://projecteuclid.org/euclid.aoas/1387823308

Examples

data(design_vec,fMRI_responses,prediction_res)

# Make example shorter - for paper example use T = ncol(fMRI_responses)
T = 156*4
fMRI_responses_sm = fMRI_responses[,1:T]
design_sm = design_vec[1:T]
permutation = rev(1:T)

prep_shuffle = prepareShuffle(design_sm,permutation)

var_explained = numeric(nrow(fMRI_responses_sm))
for (i in 1:nrow(fMRI_responses_sm)) {
    var_explained[i] = estimateShuffle(fMRI_responses_sm[i,],prep_shuffle)$effect
}

plot(var_explained, pmax(prediction_res,0)^2,
		    xlim = c(0,0.7), ylim = c(0,0.7), 
		    xlab = "Explainable variance", ylab = "Corr^2") 
abline(0,1,col=4)