Nonparametric mixture modeling for mapping of brain activation using functional MRI data
Cordes D, Nandy R
University of Washington
The hypothesis based approach to the analysis of fMRI data depends solely on the null hypothesis of non-activation. There is usually no specific alternate hypothesis primarily due to the difficulty in parametrically estimating the distribution of the observed test statistic under the hypothesis of activation. Previous approaches to this problem made several questionable assumptions. A novel nonparametric method is introduced in this abstract to formulate a mixture model by which the distribution of the test statistic under activation can be estimated. The statistical power of the test can also be accurately estimated using this method.