FANOVA models, with correlated error term, in the statistical analysis of fMRI data

J. Álvarez Liébana, M. D. Ruiz Medina

The statistical analysis of functional magnetic resonance neuroimaging
(fMRI) data has led to a substantial improvement in the field of neurology. The
investigation of the temporal evolution of the neuronal response to a stimulus usually
requires the analysis of fMRI data correlated in space and time. Functional Analysis of
Variance (FANOVA) is then considered here, from a correlated sequence of functional
data with rectangular support (i.e., from a temporal correlated sample of functional
MRI data). Specifically, a multivariate Hilbert-valued fixed effect model, with
correlated functional error term, is fitted, considering the RKHS quadratic loss
function. The optimality of the projection methodology adopted is analyzed, regarding
the dimension reduction problem. The results obtained are compared with those ones
derived by applying the FMRISTAT software, available at
http://www.math.mcgill.ca/keith/fmristat/.

Palabras clave: Functional analysis of variance, Hilbert-valued multivariate fixed effect model, Hemodynamic response function, MRI and fMRI applications

L05.2 Bioestadística
5 de septiembre de 2016  11:30
0.09 - Aula de proyectos 2

S. Pérez Fernandez, N. Corral Blanco, P. Martínez-Camblor

Y. Larriba González, C. Rueda Sabater, M. A. Fernández Temprano