Changes with version 1.3. 
added multiscale test based segmentation algorithm fro signal detection
(use adaptation="segment" in fmri.smooth)

bugfix in handling residuals in fmri.lm
removed unused branches from fmri.smooth/vaws3D
fixed handling of ladjust in ...
fixed display of bandwidth in case of lkern="Gaussian"
reduced signal in demo

mask is now forced to be connective 

Changes with version 1.1-1 

New functionality:

function ngca to perform non-Gaussian component analysis 

Changes in existing functions:

fmri.smooth: An additional option "Plateau" for skern and lkern 
allows to specify a kernel that is 1 for arguments in (0,.3), decreases
linearly in (.3,1) and equals 0 otherwise. 
Defaults for lkern and skern have been changed in 
to skern="Plateau" and lkern="Gaussian". Defaults for the parameter lambda
have been readjusted internally to meet a propagation condition with the new default kernels. This may lead to slightly different (more reliable) results.

Errors in demo(fmri) have been corrected.

